Wireless Networks

Haipeng Yao
Chunxiao Jiang
Yi Qian

Developing
Networks
using Artificial
Intelligence
Wireless Networks

Series editor
Xuemin Sherman Shen
University of Waterloo, Waterloo, ON, Canada
More information about this series at http://www.springer.com/series/14180
Haipeng Yao • Chunxiao Jiang • Yi Qian

Developing Networks using

Artificial Intelligence

123
Haipeng Yao Chunxiao Jiang
Sch of Info & Communication Engg Tsinghua Space Center
Beijing University of Posts and Telecomm Tsinghua University
China, Beijing, China Beijing, Beijing, China

Yi Qian
Dept of Electrical & Computer Engg
University of Nebraska-Lincoln
Omaha, NE, USA

ISSN 2366-1186 ISSN 2366-1445 (electronic)

Wireless Networks
ISBN 978-3-030-15027-3 ISBN 978-3-030-15028-0 (eBook)
https://doi.org/10.1007/978-3-030-15028-0

Library of Congress Control Number: 2019935167

© Springer Nature Switzerland AG 2019

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

With the software-defined networking (SDN), network function virtualization

(NFV), and fifth-generation wireless systems (5G) development, the global net-
work is undergoing profound restructuring and transformation. However, with the
improvement of the flexibility and scalability of the networks, the ever-increasing
complexity of networks makes effective monitoring, overall control, and optimiza-
tion extremely difficult. Recently, adding intelligence to the control plane through
AI and ML becomes a trend and a direction of network development.
In this book, we apply different machine learning approaches to investigate
solutions for intelligent monitoring, overall control, and optimization of networks.
We focus on four scenarios of successfully applying machine learning in network
space.
Intelligent Network Awareness In the network, different applications produce
various traffic types with diverse features and service requirements. Therefore, in
order to better manage and control networking, the intelligent awareness of network
traffic plays a significant role. In Chap. 3, we discuss the main challenge of network
traffic intelligent awareness and introduced several machine learning-based traffic
awareness algorithms, such as traffic classification, anomaly traffic identification,
and traffic prediction.
Intelligent Network Control Finding the near-optimal control strategy is the most
critical and ubiquitous problem in a network. However, the traditional works on
the control plane are largely relied on a manual process in configuring forwarding,
which cannot be employed for nowadays network condition. To address this
issue, several artificial intelligence approaches for self-learning control strategies
in networks are introduced in Chap. 4.
Intelligent Network Resource Allocation Resource management problems are
ubiquitous in the networking field, such as job scheduling, bitrate adaptation in
video streaming, and virtual machine placement in cloud computing. Compared
with the traditional with-box approach, ML methods are more suitable to solve

v
vi Preface

the complexity of network resource allocation problems. Part of these works is

introduced in Chap. 5.
Intent-Based Networking Management Future networks need to capture busi-
ness intent and activate it network-wide in order to bridge the gap between
the business needs and the network deliveries. Therefore, in this book, semantic
comprehension function is introduced to the network to understand the high-level
business intent. Part of these works is introduced in Chap. 6.
The topic of AI-driven networking is quite hot in current academia. We can
now find many books on this topic. The following two features make this book a
unique source for students and researchers. This book presents a four-tier network
architecture named NetworkAI, which is applying AI and ML to deal with different
level challenges of today’s networking. This book presents a formalized analysis
of several up-to-date networking challenges like virtual network embedding, QoS
routing, etc. Some machine learning methods were introduced to effectively solve
these challenges. These successful cases show how the ML can benefit and
accelerate the network development.
Overall, this book aims at giving a comprehensive discussion on the motivation,
problem formulation, and research methodology on applying machine learning in
the future intelligent network. Although we made an earnest endeavor for this book,
there may still be errors in the book. We would highly appreciate if you contact us
when you find any.

Beijing, China Haipeng Yao

Beijing, China Chunxiao Jiang
Omaha, NE, USA Yi Qian
Acknowledgments

Thanks to all the collaborators who have also contributed to this book: Tianle Mai,
Mengnan Li, Chao Qiu, Peiying Zhang, Yaqing Jin, and Yiqi Xue.

vii
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Background. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Overview of SDN and Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Software Defined Networking (SDN) . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.2 Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Related Research and Development. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.1 3GPP SA2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.2 ETSI ISG ENI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.3 ITU-T FG-ML5G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4 Organizations of This Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2 Intelligence-Driven Networking Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1 Network AI: An Intelligent Network Architecture for
Self-Learning Control Strategies in Software Defined Networks. . . . . 13
2.1.1 Network Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.2 Network Control Loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.1.3 Use Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.1.4 Challenges and Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3 Intelligent Network Awareness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1 Intrusion Detection System Based on Multi-Level
Semi-Supervised Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1.1 Proposed Scheme (MSML) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.1.2 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2 Intrusion Detection Based on Hybrid Multi-Level Data Mining . . . . . 50
3.2.1 The Framework of HMLD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.2.2 HMLD with KDDCUP99 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2.3 Experimental Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . . 65

ix
x Contents

3.3 Abnormal Network Traffic Detection Based on Big Data

Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.3.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.3.2 Simulation Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4 Intelligent Network Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.1 Multi-Controller Optimization in SDN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.1.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.1.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.1.3 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.2 QoS-Enabled Load Scheduling Based on Reinforcement
Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.2.1 System Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.2.2 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.2.3 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.2.4 Simulation Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.3 WLAN Interference Self-Optimization Based SOM Neural
Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.3.1 Som Network Training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.3.2 Som Neural Network Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.3.3 Network Training And Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.3.4 Power Efficiency Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
4.4 Blockchain-Based Software-Defined Industrial Internet of
Things: A Dueling Deep Q-Learning Approach. . . . . . . . . . . . . . . . . . . . . . . 127
4.4.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.4.2 Network Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.4.3 Blockchain-Based Consensus Protocol . . . . . . . . . . . . . . . . . . . . . . . 131
4.4.4 Detailed Steps and Theoretical Analysis . . . . . . . . . . . . . . . . . . . . . . 133
4.4.5 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
4.4.6 Dueling Deep Q-Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
4.4.7 Simulation Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
5 Intelligent Network Resource Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.1 Virtual Network Embedding Based on RDAM . . . . . . . . . . . . . . . . . . . . . . . . 157
5.1.1 System Model and Problem Formulation. . . . . . . . . . . . . . . . . . . . . . 158
5.1.2 RDAM Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
5.1.3 Dynamic Update of Substrate Network. . . . . . . . . . . . . . . . . . . . . . . . 165
5.1.4 Network Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
5.1.5 Training and Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
5.1.6 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Contents xi

5.2 Virtual Network Embedding Based on Policy Network and

Reinforcement Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
5.2.1 Network Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
5.2.2 Embedding Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
5.2.3 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
5.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
6 Intention Based Networking Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
6.1 CNN Based Sentence Similarity Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
6.1.1 The Proposed Model WSV-SCNN-SV . . . . . . . . . . . . . . . . . . . . . . . . 201
6.1.2 Experimental Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
6.2 A Feature Selection Method Based Text Classification System . . . . . . 215
6.2.1 Text Classification Model Based on Semantic Similarity. . . . . 217
6.2.2 Algorithm Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
6.2.3 Experiments and Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
6.3 Sematic and Statistical Information Based Text Classification . . . . . . . 225
6.3.1 Word Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
6.3.2 Merged Kernel Function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
6.3.3 Experiments and Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
7 Conclusions and Future Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
7.2 Future Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
7.2.1 New Hardware Architecture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
7.2.2 Advancing Software System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
7.2.3 Self-Adaptive Network Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
7.2.4 Promoting Machine Learning Algorithms . . . . . . . . . . . . . . . . . . . . . 248
Chapter 1
Introduction

1.1 Background

The current Internet architecture established from TCP/IP has gained huge success
and been one of the indispensable infrastructures for our daily life, economic
operation and society. However, burgeoning megatrends in the information and
communication technology (ICT) domain are urging the Internet for pervasive
accessibility, broadband connection and flexible management, which call for poten-
tial new Internet architectures. The original design tactic of the Internet, which is
“Leaving the complexity to hosts while maintaining the simplicity of network”,
leads to the almost insurmountable challenge known as “Internet ossification”: soft-
ware in the application layer has developed rapidly, and abilities in the application
layer have been drastically enriched. By contrast, protocols in the network layer
lack scalability and the core architecture is hard to modify, which means that new
functions have to be implemented through myopic and clumsy ad hoc patches in the
existing architecture. For example, the transition from IPv4 to IPv6 is difficult to
deploy in practice.
To improve the performance of the current Internet, novel network architectures
have been proposed by the research communities to build the future Internet, such
as Content Centric Networking (CCN) and Software-Defined Networking (SDN).
Specifically, the past few years have witnessed a wide deployment of software
defined networks. SDN is a paradigm, which separates the control plane from
the forwarding plane, breaks vertical integration, and introduces the ability of
programming the network. However, the work on the control plane largely relies
on a manual process in configuring forwarding strategies. With the expansion of
network size and the rapid growth of the number of network applications, current
networks have become highly dynamic, complicated, fragmented, and customized.
These requirements pose several challenges for tradition SDN.
Recently, AlphaGo’s success comes at a time when researchers are exploring
the potential of artificial intelligence to do everything from drive cars to financial

© Springer Nature Switzerland AG 2019 1

H. Yao et al., Developing Networks using Artificial Intelligence,
Wireless Networks, https://doi.org/10.1007/978-3-030-15028-0_1
2 1 Introduction

market. Therefore, we ask if machine learning can provide a feasible alternative

to a manual process for networking control and management. In other words, can
software defined networking learn to manage resource on their own by machine
learning technology? The recent success of using machine learning in many
challenging domains suggest that this aim may not be impossible. Therefore, in this
book, we focus on how machine learning can revolute networking development.

1.2 Overview of SDN and Machine Learning

In this section, we provide a more detailed view on software defined networking and
machine learning.

1.2.1 Software Defined Networking (SDN)

SDN is a new type of network architecture. Its design concept is to separate the
control plane of the network from the data forwarding plane, so as to realize the
programmable control of the underlying hardware through the software platform
in the centralized controller and realize flexible network resources. On-demand
deployment. In the SDN network, the network device is only responsible for
pure data forwarding, and can use common hardware; the operating system that
is responsible for control will be refined into an independent network operating
system, which is responsible for adapting different service characteristics, and the
network operating system and Business characteristics and communication between
hardware devices can be implemented programmatically.
As shown in Fig. 1.1, the SDN consists of three layers, forwarding plane, control
plane and application plane.
Forwarding Plane Only responsible for network forwarding, not for the lowest
infrastructure layer responsible for flow table-based data processing, forwarding,
and state collection.
Control Plane The controller centrally manages all the devices on the network.
The virtual entire network is a resource pool. The resources are flexible and
dynamically allocated according to the different needs of users and the global
network topology. The SDN controller has a global view of the network and is
responsible for managing the entire network: for the lower layer, communicating
with the underlying network through standard protocols; for the upper layer,
providing the application layer with control over the network resources through the
open interface.
Application Plane The top layer is the application layer, including various services
and applications; the middle control layer is mainly responsible for processing the
1.2 Overview of SDN and Machine Learning 3

Fig. 1.1 The architecture of SDN

data plane resources, maintaining the network topology, state information, etc.; DN
application layer through the programming layer provided by the programming
interface The underlying device is programmed to open the control of the network to
the user, and to develop various business applications based on the above, to achieve
rich and colorful business innovation.
In addition, SDN includes two interfaces:
Southbound Interface The southbound interface is the channel through which
the physical device and the controller transmit signals. The related device status,
data flow entries, and control commands need to be communicated through the
southbound interface of the SDN to implement device control.
Northbound Interface The northbound interface is an interface that is open to
the upper layer service application through the controller. The purpose is to enable
the business application to conveniently call the underlying network resources and
capabilities. The service is directly applied to the service application, and the design
needs to be closely related to the service application requirements.
The traditional network adopts a three-layer distributed network, a core layer,
an aggregation layer and an access layer. Without a unified central control node,
devices on the network learn the direct routing information advertised by other
devices, so that each device decides how to forward it. This makes it impossible to
control traffic from the perspective of the entire network. The typical architecture of
SDN is divided into three layers: Forwarding plane, Control plane, and Application
plane.
Therefore, compared with traditional networks SDN has the following advan-
tages: first, the device hardware is normalized. The hardware only focuses on
forwarding and storage capabilities and decouples from the service features. It can
4 1 Introduction

be implemented using a relatively inexpensive commercial architecture. Second, the

intelligence of the network is all implemented by software. The types and functions
of the network devices are determined by software configuration. The operation
control and operation of the network is performed by the server as a network
operating system (NOS). Third, the service response is relatively faster, and various
network parameters such as routing, security, policy, QoS, traffic engineering, etc.
can be customized, and configured in real time to the network, and the time for
opening specific services will be shortened.
The essence of SDN is to enable users/applications to fully control the behavior
of the network through software programming, to make the network software and
agile. Its typical features are as followed. Control and forwarding are separated. The
forwarding plane consists of controlled forwarding devices. The forwarding mode
and service logic are controlled by the control application running on the separated
control plane. An open interface between the control plane and the forwarding
plane. SDN provides an open programmable interface to the control plane. In this
way, the control application only needs to focus on its own logic, without having
to pay attention to the underlying implementation details. Logically centralized
control. The logically centralized control plane can control multiple forwarding
plane devices, that is, control the entire physical network, thereby obtaining a global
network state view, and implementing optimal network control according to the
global network state view.
Based on the above points, SDN has the following advantages for the network
as followed. First, SDN provides more flexibility for the use, control, and how to
generate revenue for the network. Second, SDN has accelerated the introduction of
new services. Network operators can deploy related functions through controlled
software without having to wait for a device provider to add a solution to their
proprietary device as before. Third, SDN reduces the operating expenses of the
network and reduces the error rate. The reason is that the network is automatically
deployed and the operation and maintenance fault diagnosis is realized, and the
manual intervention of the network is reduced. Fourth, SDN helps to realize the
virtualization of the network, thus realizing the integration of computing and storage
resources of the network, and finally, the control and management of the entire
network can be realized by a simple combination of software tools. Fifth, SDN
makes the network and even all IT systems better oriented to business goals.

1.2.2 Machine Learning

Over the past few decades, machine learning techniques have attracted lots of
attention from both academia and industry. Machine learning was proposed in the
late 1950s as a key approach for Artificial Intelligence (AI) for the first time. The
classical definition of ML can be described as “The development of computer
models for training data that provides solutions to the problems of knowledge
1.2 Overview of SDN and Machine Learning 5

Fig. 1.2 Evolutions of machine learning techniques

acquisition and enhances the performance of developed systems without being

explicitly programmed to”.
We will have a look at machine learning’s evolutions from the 1950s until now
in Fig. 1.2. The first appearance of ML happened in the 1950s, and some early
researches had been done from 1950 to 1980. With the emerging of data mining,
there was a phenomenal burst of research and development in ML. Data mining
is a group of algorithms about extracting the useful and unknown knowledge from
datasets. Before this, ML was only used in classifying and predicting based on the
known properties extracted from the training data, which needed the appropriate
problem formulations, while data mining was used to discover the unknown
properties from datasets. Data mining does not need to build an appropriate problem
formulation but focuses on discovering valuable knowledge. During this period,
mobile data communication technologies (e.g., LTE and LTE-A) developed rapidly,
and the users’ data exploded in both volume and speed, which enabled to be used
to understand the performances and interests of users. Thus, the first practical
applications appeared, such as advertisements and recommendations toward users’
behaviors.
From the 2010s until now, an even more intelligent branch of ML, called as
deep learning, has attracted the whole world’s attention. If we need the novel
insights from data to construct the computational models aiming at the high-
level data abstractions, one tool for this is deep learning, hierarchical learning.
Deep learning enables to use the unlabeled data to discover the intermediate or
abstract representations in a hierarchical manner with deep neural networks. The
higher-level features are defined based on the lower-level features to improve the
classification or modeling results. Especially, there is a breakthrough when the
distributed version of AlphaGo defeated the European Go champion Fan Hui for
the first time. AlphaGo utilizes the Monte Carlo tree search algorithm and artificial
neural networks to find the optimal moves.
6 1 Introduction

Fig. 1.3 Various machine learning algorithms and their corresponding classification and applica-
tions

After reviewing the evolution of machine learning, some items about ML

algorithms will be presented. Existing machine learning techniques can be classified
into supervised learning, unsupervised learning, and reinforcement learning. In
supervised learning, algorithms need the labeled training datasets to construct the
system models about the relations among inputs, outputs, and parameters. On
the contrary, unsupervised learnings are not provided with labels and their main
applications are to classify the sample sets into the different clusters based on
the investigations of similarity between inputs. The third category, reinforcement
learning, is another area of ML, where the agent takes actions in an environment
to maximize the cumulative reward. Finally, in order to understand ML algorithms
systematically, various ML algorithms, their corresponding classification and appli-
cations are illustrated in Fig. 1.3.

1.2.2.1 Supervised Learning

Supervised learning is the process of adjusting the parameters of a classifier to the

desired performance using a set of samples of a known class. Supervised learning
is a machine learning task that infers a function from the labeled training data.
The training data includes a set of training examples. In supervised learning, each
instance consists of an input object (usually a vector) and the desired output value
(also known as a supervisory signal). The supervised learning algorithm analyzes
1.2 Overview of SDN and Machine Learning 7

the training data and produces an inferred function that can be used to map out new
instances. An optimal solution would allow the algorithm to correctly determine
the class labels of those invisible instances. This requires learning algorithms to be
formed in a “reasonable” way from one training data to invisible.
The training set requirements for supervised learning include input and output,
and can also be said to be features and goals. The goal of the training set is to
be labeled (scalar) by the person. Under supervised learning, the input data is called
“training data” and each set of training data has a clear identification or result. When
establishing a predictive model, supervised learning establishes a learning process,
compares the predicted results with the actual results of the “training data”, and
continuously adjusts the predictive model until the predicted outcome of the model
reaches an expected accuracy.
Supervised learning is divided into classification algorithms and regression
algorithms: For classification, the target variable is the category to which the sample
belongs. In the sample data, the characteristics of each sample are included, for
example, the training samples of various data input to the model are generated,
and the data of one person is input to determine whether or not the patient has
the data. As a result of cancer, the results must be discrete, with only “yes” or
“no.” For regression, regression is for predicting, for example, a training sample that
inputs various data of a person into a model, and produces a result of “inputting one
person’s data and judging the future economic ability of the person 20 years later”,
and the result is continuous. Often get a regression curve. When the input arguments
are different, the output dependent variables are not discretely distributed. The
difference between classification and regression is that the target variable of the
classification is discrete, and the target variable of the regression is continuous
numerical.

1.2.2.2 Unsupervised Learning

Compared with supervised learning, the unsupervised learning training set has
no artificially labeled results. In unsupervised learning, data is not specifically
identified, and the learning model is used to infer some of the inherent structure
of the data. Common application scenarios include learning of association rules
and clustering. Common algorithms include the Apriori algorithm and the K-Means
algorithm. The goal of this type of learning is not to maximize the utility function,
but to find the approximate point in the training data. Clustering often finds fairly
good visual classifications that match hypotheses. For example, aggregates based
on demographics may form a rich aggregate in one group, as well as other poor
aggregates.
Unsupervised learning seems very difficult: the goal is that we don’t tell the
computer what to do, but let it (computer) learn how to do something. There are
two general ideas for unsupervised learning: The first idea is to not specify a clear
classification when guiding the Agent but to adopt some form of incentive system
when it succeeds. It should be noted that this kind of training is usually placed in
8 1 Introduction

the framework of decision-making problems because its goal is not to produce a

classification system but to make the decision to maximize returns. This kind of
thinking is a good summary of the real world, Agent can motivate those correct
behaviors and penalize other behaviors.
Commonly used unsupervised learning algorithms mainly include principal
components analysis methods such as PCA, equidistant mapping method, locally
linear embedding method, Laplacian feature mapping method, black plug local
linear embedding method and local tangent spatial permutation method. A typical
example of unsupervised learning is clustering. The purpose of clustering is to bring
together similar things, and we don’t care what this class is. Therefore, a clustering
algorithm usually only needs to know how to calculate the similarity and can start
working.

1.2.2.3 Reinforcement Learning

Reinforcement learning is a reward-guided behavior in which the agent learns in

a “trial and error” manner, and the goal is to maximize the reward of the agent
through the interaction with the environment. Reinforcement learning is different
from the supervised learning in connected learning. It is mainly reflected in the
teacher’s signal. The enhanced signal provided by the environment in reinforcement
learning is to evaluate the quality of the action (usually a scalar signal), rather
than telling the reinforcement learning system RLS (reinforcement learning system)
how to generate the correct action. Because the external environment provides very
little information, RLS must learn on its own experience. In this way, RLS gains
knowledge in an action-evaluation environment and improves the action plan to
adapt to the environment.
Reinforcement learning is developed from theories of animal learning, parameter
disturbance adaptive control, etc. The basic principles are: If a certain behavioral
strategy of the Agent leads to a positive reward (enhanced signal), then the trend of
the Agent to generate this behavioral strategy will be strengthened. The goal of the
Agent is to find the optimal strategy in each discrete state to maximize and reward
the desired discount.
Reinforcement learning regards learning as a process of tentative evaluation.
The Agent selects an action for the environment. After the environment accepts
the action, the state changes, and at the same time, an enhanced signal (a reward
or punishment) is sent to the Agent. The agent re-intensifies the signal and the
current state of the environment. Choosing the next action, the principle of choice
is to increase the probability of being positively strengthened (award). The selected
action not only affects the immediate enhancement value but also affects the state
of the environment at the next moment and the final enhancement value.
1.3 Related Research and Development 9

1.3 Related Research and Development

1.3.1 3GPP SA2

3GPP SA2 set up a research project “Study of enablers for Network Automation for
5G (eNA)” on 5G network intelligence at the Hangzhou conference in May 2017.
The project belongs to 3GPP Rel 16, SA2 has officially discussed the project at the
Gothenburg meeting in January 2018.
The project background of the project is that 3GPP SA2 introduces Network Data
Analysis Function (NWDAF) in Rel 15’s 5G core network. At present, the main
application scenario of the function is a slice network, which provides network slice
state analysis results to the policy control function and the network slice selection
function by automatically analyzing the network data related to the network slice.
On the other hand, in the 5G network architecture research of Rel 15, SA2
introduced some new requirements, such as on-demand mobility management, non-
standardized QoS, traffic offloading and offloading. Without network data analysis,
it is difficult to actually deploy and implement these requirements. Therefore, in
order to make the 5G network more flexible and intelligent, Huawei has led the
establishment of the eNA project in SA2.
The research goal of the eNA project is to collect and analyze network data
through NWDA, generate analysis results, and then use the analysis results for net-
work optimization, including customized mobility management, 5G QoS enhance-
ment, dynamic traffic grooming and offload, and user plane functions. Select, based
on the service usage of the UE, traffic policy routing and service classification.

1.3.2 ETSI ISG ENI

Recently, the European Telecommunications Standards Institute ETSI officially

approved the establishment of a new industry standard group ENI (Experiential
Networked Intelligence). The working group’s goal is to align the scenarios and
requirements of the network intelligence, define a closed-loop control model based
on “observation-judgment-decision-action” and simulate human brain decision-
making, and continuously optimize the network environment and continuously
optimize decision-making results. Means to effectively cope with complex network
management and control challenges and improve network operation management
efficiency and experience.
The research scope of the working group includes analyzing the operational
and operation and maintenance requirements of operators’ traditional networks
and SDN/NFV networks and guiding operators to build an architecture including
adaptive sensing, flexible policy definition support, and intelligent decision-making
and execution. The architecture should fully support operators’ flexible business
strategies and intelligent, self-optimizing, and autonomous smart network concepts.
10 1 Introduction

At the same time, the architecture will also promote the development of technologies
such as telemetry remote sensing, big data collection and management, and machine
learning algorithms to support intelligent analysis and decision making. The unified
strategy model is one of the core research technologies of network artificial
intelligence.
The current ENI work process is divided into two phases: the first phase defines
and describes the use cases and requirements, and makes them agree, analysing the
gap between use cases and requirements and liaising with relevant standards groups.
In the second phase, ENI will define the corresponding network architecture based
on the results of the first phase, and design the use cases and requirements defined
in the first phase. Currently, ENI has launched four related technical manuscript
projects: use cases, requirements, context-aware strategy management gap analysis,
terminology. At present, the drafting of these four manuscripts has come to an end,
and the manuscript will be released after being reviewed. At the same time, ENI
launched the architecture project in January 2018.

1.3.3 ITU-T FG-ML5G

During the ITU-T SG13 meeting in November 2017, representatives of business

experts from different countries such as Germany, South Korea, China, Tunisia,
and Africa proposed the establishment of a machine learning-network focus group,
which was officially approved by the SG13 plenary discussion. The official name of
the focus group is Machine Learning for Future Networks including 5G, referred to
as FG-ML5G.
The FG-ML5G is a working platform open to both ITU and non-ITU members.
The goal is to analyze how to apply machine learning to future networks, especially
5G networks, to improve network performance and user experience. Specific work
includes analyzing and collaborating with current industry standards organizations;
researching machine learning scenarios, potential needs, architectures, and specific
interfaces, protocols, algorithms, data structures, and personal information protec-
tion in future networks; analyzing machine learning Impact on autonomous network
control and management.
FG-ML5G is the focus group under SG13. It has three working groups internally
to study application scenarios and business requirements, data format, machine
learning technology, and network architecture. The research period is 1 year, and it
can be determined by SG13 if necessary. Whether to postpone. The research report
and draft standard of the FG-ML5G output can be used as input for subsequent
research on SG13.
1.4 Organizations of This Book 11

1.4 Organizations of This Book

This book is organized as depicted in Fig. 1.4. We first propose the concept of Net-
workAI, a novel paradigm that applying machine learning to automatically control
a network. NetworkAI employs reinforcement learning and incorporates network
monitoring technologies such as the in-band network telemetry to dynamically
generate control policies and produces a near optimal decision. We employ the SDN
and INT to implement a network state upload link and a decision download link to
accomplish a closed-loop control of a network and build a centralized intelligent
agent aiming at learning the policy by interaction with a whole network.
Then, we discuss the possible machine learning methods for network awareness.
With the rapid development of compelling application scenarios of the networks,
such as 4K/8K, IoT, it becomes substantially important to strengthen the manage-
ment of data traffic in networks. As a critical part of massive data analysis, traffic
awareness plays an important role in ensuring network security and defending traffic
attacks. Moreover, the classification of different traffic can help to improve their
work efficiency and quality of service (QoS).
Furthermore, we discuss how machine learning can achieve network auto-
matically control. Finding the near-optimal control strategy is the most critical
and ubiquitous problem in a network. Examples include routing decision, load
balancing, QoS-enable load scheduling, and so on. However, the majority solutions
of these problems are largely relying on a manual process. Therefore, to address this

natural
language
understandi
Chapter 6: Intenon based
ng networking management
natural
language Intent-based networking
Users

Resource management Chapter 5: Networking

Traffic
Traffic Engineering
Resource management
QoS\QoE
VNE
Traffic
QoS\QoE Chapter 4: Roung and Traffic
Network State Machine learning Routing
Engineering
Network Analytics platform

Network Network Chapter 3: Network

Awareness SDN Controller Control Awareness

Chapter 2: Network
Architecture

Fig. 1.4 Book organization

12 1 Introduction

issue, in Chap. 4, we apply several artificial intelligence approaches for self-learning

control strategies in networks.
In addition, resource management problems are ubiquitous in the networking
field, such as job scheduling, bitrate adaptation in video streaming and virtual
machine placement in cloud computing. In Chap. 5, we focus on how machine
learning can optimal resource management decision.
Finally, in order to reduce network complexity and improves automation by
eliminating manual configurations, we discuss the intent-based networking which
allows a user or administrator to send a simple request—using natural language—
to plan, design and implement/operate the physical network. For example, an IT
administrator can request improved voice quality for its voice-over-IP application,
and the network can respond. For intent-based networking, the translation and
validation system take a higher-level business policy (what) as input from end users
and converts it to the necessary network configuration (how) by natural language
understanding technology. To achieve intent-based networking, in Chap. 6, we focus
on how machine learning can technology can be used in the natural language
understanding in translation and validation system.

1.5 Summary

In this chapter, we mainly introduce the background. We first introduce the moti-
vation of this book. Based on the motivation, we propose the coordinated network
architecture and discuss the key technologies and challenges in the architecture.
Then, the related research and development are provided. Finally, we give the
organizations of this book.
Chapter 2
Intelligence-Driven Networking
Architecture

Network architecture provides a full picture of the established network with a

detailed view of all the resources accessible. The architecture of software defined
network facilitates separation of the control plane from the forwarding plane.
However, the work on the control plane largely relies on a manual process in
configuring forwarding strategies. To address this issue, in this chapter, we propose
NetworkAI, a new network architecture exploiting software-defined networking,
network monitor technologies and reinforcement learning technologies for con-
trolling networks in an intelligent way. NetworkAI implements a network state
upload link and a decision download link to accomplish a closed-loop control of
the network and builds a centralized intelligent agent aiming at learning the policy
by interacting with the whole network.

2.1 Network AI: An Intelligent Network Architecture for

Self-Learning Control Strategies in Software Defined
Networks

Recently, Software Defined Networking (SDN) has received a large amount of

attention from both academia and industry. The SDN paradigm decouples control
plane from data plane and provides a logically-centralized control plane, wherein
the network intelligence and state are logically centralized, and the underlying
network infrastructure is abstracted for the network applications and services [1, 7].
This logically-centralized control mechanism provides the efficient use of network
resources for network operators and the programmability brought by SDN simplifies
the configuration and management of networks. Therefore, network operators can
easily and quickly configure, manage and optimize network resource in SDN
architecture [3]. However, even though a separation of the control plane from the
forwarding plane facilitates large scale and dynamic networks, a challenge that

© Springer Nature Switzerland AG 2019 13

H. Yao et al., Developing Networks using Artificial Intelligence,
Wireless Networks, https://doi.org/10.1007/978-3-030-15028-0_2
14 2 Intelligence-Driven Networking Architecture

remains un-tackled is that the work on the control plane relies heavily on a manual
process in configuring forwarding strategies.
Finding the near-optimal control strategy is the most critical and ubiquitous
problem in a network. The majority of approaches to solve this problem today
usually adopted the white-box approaches [5, 6]. With the expansion of network
size and the rapid growth of the number of network applications, current networks
have become highly dynamic, complicated, fragmented, and customized. These
requirements pose several challenges when applying these traditional white-box
algorithms [4]. Specifically, a white-box approach generally requires an idealized
abstraction and simplification of the underlying network; however, this idealized
model often poses difficulties when dealing with a real complex network environ-
ment. In addition, the white-box method presents poor scalability under different
scenarios and applications.
Owing to the success of Machine Learning (ML) related applications, such as
robotic control, autonomous vehicles, and Go [8], a new approach for network
control through ML has emerged. This new networking paradigm using is firstly
proposed by Mestres et al., which is referred to as Knowledge-Defined Networking
(KDN) [9]. However, KDN and some other similar works [10, 12] just proposed
some concept, no detail was described in these papers and no actual work has been
implemented.
In this part, we propose NetworkAI, an architecture exploiting software-defined
networking, network monitor technologies (e.g., traffic identification, In-band Net-
work Telemetry(INT)), and reinforcement learning technologies for controlling net-
works in an intelligent way. NetworkAI implements a network state upload link and
a decision download link to accomplish a close-loop control of network and builds
a centralized intelligent agent aiming at learning the policy by interacting with the
whole network. The SDN paradigm decouples control plane from data plane and
provides a logically-centralized control to whole underlying network. Some new
network monitor technologies, such as In-band Network Telemetry(INT), [2, 11, 13]
can achieve millisecond uploading of the network state and provide real-time packet
and flow-granularity information to a centralized platform [14]. In addition, a
network analytical platform, such as the PNDA [15], provides big data processing
services via some technologies such as Spark and Hadoop. SDN and monitor
technologies offer a completely centralized view and control to build the interaction
framework of a network, thus enabling the application of ML running in a network
environment to address the network control issues.
NetworkAI applies DRL to effectively solve real-time large-scale network
control problems without relying on too much manual process and any assumptions
of the underlying network. RL involves agents that learn to make better decisions
from experiences by interacting with the environment [13, 17]. During training, the
intelligent agent begins with no prior knowledge of the network task at hand and
learns by reinforcement based on its ability to perform a task. Particularly, with the
development of deep learning (DL) techniques, the success of combined application
of RL and DL to large-scale system control problems (such as GO [8] and playing
games) proves that the deep reinforcement learning (DRL) algorithm can deal with
2.1 Network AI: An Intelligent Network Architecture for Self-Learning. . . 15

the complicated system control problem. DRL represents its control policy as a
neural network that can transfer raw observations (e.g., delay, throughput, jitter) to
the decision [16]. Deep learning (DL) can effectively compress the network state
space thus enabling RL to solve large-scale network decision-making problems that
were previously found difficult in handling high-latitude states and motion space.
In NetworkAI, the SDN and new network monitor technologies are employed
herein to construct a completely centralized view and control for geographical
distributed network and build a centralized intelligent agent to generate a network
control policy via DRL. The NetworkAI can intelligently control and optimize a
network to meet the differentiated network requirements in a large-scale dynamic
network.
Different from the traditional white-box approaches, this part proposes a new
network paradigm (i.e. NetworkAI) in applying ML to solve network control
problem. The main contribution of this section can be summarized as follows:
• We employ the SDN and INT to implement a network state upload link and
a decision download link to accomplish a close-loop control of a network and
build a centralized intelligent agent aiming at learning the policy by interaction
with a whole network.
• We apply DRL to effectively solve real-time large scale network control prob-
lems without too much manual process and any assumptions of underlying
network, where the DRL agent can produce a near-optimal decision in real time.

2.1.1 Network Architecture

In this subsection, we firstly introduce the NetworkAI architecture and elaborate

how it operates. The model of the NetworkAI is shown in Fig. 2.1. This model
consists of three planes called the forwarding plane, the control plane, and the AI
plane. And then, we will give a detailed description of each layer.

2.1.1.1 Forwarding Plane

The function of the forwarding plane is forwarding, processing, and monitoring data
packets [18]. The network hardware, which is composed of line-rate programmable
forwarding hardware, only focuses on simply data forwarding without embedding
any control strategies. The control rules are issued by the SDN controller via
southbound protocols such as OpenFlow [23] or P4 [24]. When a packet comes
into the node, it will be forwarded and processed according to these rules. Besides,
there are some monitoring processes embedded in nodes. The network monitor data
will be collected and sent to the analytical platform. Thus, it can offer complete
network state information to facilitate the AI plane to make decisions.
16 2 Intelligence-Driven Networking Architecture

Fig. 2.1 The NetworkAI architecture

2.1.1.2 Control Plane

The function of the control plane is to connect the AI plane and the forwarding
plane. This plane provides abstractions for accessing lower-level geographical
distributed forwarding plane and pools the underlying resources (such as link
bandwidth, network adapter, and CPU capacity) to the AI plane. The function of the
SDN controller is to manage the network through standard southbound protocols
and interact with the AI plane through the northbound interfaces. This logically-
centralized plane eases the burden of the network control problem imposed by a
geographical distributed network. Thus, the policies generated by the AI plane can
be quickly deployed into the network.

2.1.1.3 AI Plane

The function of the AI plane is to generate policies. In the NetworkAI paradigm, the
AI plane takes advantage of SDN and monitor techniques to obtain a global view
and control of the entire network. The AI agent learns the policy through interaction
with the network environment. While learning the policy is a slow process, the
network analytical platform provides a big data storage and computing capacity.
Fundamentally, the AI agent processes the network state collected by the forwarding
2.1 Network AI: An Intelligent Network Architecture for Self-Learning. . . 17

plane, then transforming the data to a policy through RL and using that policy to
make decisions and optimization.

2.1.2 Network Control Loop

In the NetworkAI architecture, we design the network state upload link and
the decision download link to accomplish a close-loop control of network. The
NetworkAI architecture operates with a control loop to provide an interactive
framework for a centralized agent to automatically generate strategies. Now, in this
subsection, we will detail how the NetworkAI architecture implement a control
loop of the whole network and how the intelligent agent learns the policy by RL
approach.

2.1.2.1 Action Issue

In traditional distributed networks, the control plane of the network node is closely
coupled with the forwarding plane which has only partial control and view over the
complete network. This partial view and control can lead to no global convergence
of learning result. The AI agent need to continually converge to a new result when
network state changed which will lead to a bad performance in real-time control
problem. For purpose of achieving global optimum, the controlling and managing
of the whole network is premise. SDN is a paradigm which separates the control
plane from the forwarding plane and therefore breaks the vertical integration. The
SDN controller treats the entire network as a whole. In this manner, the SDN acts
as a logical-centralized agent to control the whole network. The SDN controller
issues a control action through an open and standard interface (e.g., OpenFlow, P4).
These open interfaces enable the controller to dynamically control heterogeneous
forwarding devices, which is difficult to achieve in traditional distributed networks.
As demonstrated in Fig. 2.2, the agent can issue control actions to the forwarding
plane via southbound protocols according to the decisions made at the AI plane, the
network node at the forwarding plane operates based on the updated rules imposed
by the SDN controller [9]. In this manner, we realized the global controllability of
the entire network.

2.1.2.2 Network State Upload

In the SDN network architecture, the controller can send an action decision
to the underlying network with an aim to acquire a complete network control.
Furthermore, obtaining the complete real-time view of the whole network is also
crucial to make near-optimal decisions. The most relevant data that should be
collected is network state information and traffic information. To this end, we
18 2 Intelligence-Driven Networking Architecture

Fig. 2.2 The action issuing process

designed the upload link to access fine-grained network and traffic information. In
this subsection, we will introduce how the NetworkAI architecture achieves network
state upload.
1. Network Information: Network information mainly refers to the status of the
network device (information below the L2 layer), including the network physical
topology, hop latency, and queue occupancy. In our architecture, we borrow in-
band telemetry technology to achieve fine-grain network monitoring.
Obtaining fine grained network monitoring data of dynamic networks is a
concern of NetworkAI. The traditional monitor technologies are commonly
based on out-band approaches. In this way, monitoring traffic is sent as dedicated
traffic, independent from the data traffic (“probe traffic”), such as SNMP,
synthetic probes. These methods bring to much probe traffic in networking and
overhead computation to control plane in large-scale dynamic networking which
extremely degrades the performance of real-time controlling.
In-band network telemetry is a framework designed to allow the collection
and reporting of network state by the data plane, without requiring intervention
or computation by the control plane. The core idea of INT is to write the network
status into the header of a data packet to guarantee the granularity of monitoring
the network at the packet level [25]. The telemetry data is straightforward adding
to the packet. Therefore, the end-to-end monitoring data can be retrieved from the
forwarding node through Kafka or IPFIX directly to AI plane’s big data platform
without intervening by the control plane.
The INT monitoring technology model is illustrated in Fig. 2.3. From the
Fig. 2.3, we can see that a source node embeds instructions in the packets, listing
the types of network information needs to be collected from the network elements
(e.g., hop latency, egress port TX link utilization, and queue occupancy). Each
2.1 Network AI: An Intelligent Network Architecture for Self-Learning. . . 19

Fig. 2.3 The collection process of network information

network element inserts the requested network state in the packet as it traverses
the network. When the packet is sent to the INT Sink, the load data is sent to the
user and the telemetry data is sent to the network analytical plane.
Data collection can be realized based on the actual traffic. INT provides
the ability to observe and collect real-time and end-to-end network information
across physical networks. In addition, the mechanism of INT vanishes the over-
head communication of probe traffic and overhead computation of control plane.
Due to borrowing from INT technology, the AI plane can obtain millisecond fine
grain network telemetry data, which gives the possibility to react network in time.
2. Traffic Information: Traffic information mainly include service-level infor-
mation (e.g., QoS/QoE), anomaly traffic detection information(e.g., elephant
flow), etc. In network, different applications produce various traffic types with
diverse features and service requirements. In order to better manage and control
networking, the identification of network traffic plays a significant role [26].
For instance, elephant flow is an extremely large continuous flow established
by a TCP (or other protocol) flow [27]. Elephant flows can occupy network
bandwidth and bring seriously congestion to the network. Therefore, it is of great
significance for AI plane to detect or even predict the elephant flow in time and
take the necessary action to avoid network congestion.
In our architecture, as illustrated in Fig. 2.4, several monitor processes
embedded in some nodes to transfer the raw traffic data (e.g., flow granularity
data, relevant traffic feature, and Deep Packet Inspection (DPI) information) to
the traffic information via data mining methods, such as traffic classification and
traffic anomaly detection [28, 30]. Then, the traffic information will be upload to
the network analytical plane to assist AI plane in decision making.
20 2 Intelligence-Driven Networking Architecture

Fig. 2.4 The collection process of traffic information

2.1.2.3 Policy Generation

With the aim to apply ML method to realize an intelligent network control, we have
already constructed the interaction framework between an AI agent and the network
environment. In this part, we will describe how to use ML method to generate
the network policy. The machine learning methods contain three approaches,
supervised learning, unsupervised learning and reinforcement learning. Compared
to supervised and unsupervised learning, reinforcement learning is more suitable for
close-loop control problems. In particular, with the development of DL, the success
of combining DL and RL for applications in decision-making domains (Playing
Atari with Deep Reinforcement Learning by DeepMind at NIPS 2013 and the 2016
Google AlphaGo success on Go) demonstrates that DRL can effectively solve large-
scale system control problems. Thus, we apply RL method to solve the large-scale
network control problem.
RL based learning tasks are usually described as a Markov decision process,
as shown in Fig. 2.5. At each step, an agent observes the current state st from the
network environment and the agent takes an action at according to a policy π(a|s).
Following the action, the network environment transfer to state st + 1 and the
agent observes a reward signal rt from environment. The goal of the reinforcement
learning is to obtain the optimal behavior policy that maximizes the expected long-
term reward. Specifically, in the network scenario, the state is represented by the
network state and flow information, the action is by network behaviors (e.g., CDN
selection, routing selection) and the reward is based on the optimal target.
2.1 Network AI: An Intelligent Network Architecture for Self-Learning. . . 21

Fig. 2.5 The reinforcement learning process

The RL agent uses a state-action value function Q(s, a) to measure the actions
expected for a long-term reward on the state s. Starting from a random Q-function,
in q-learning algorithm, the agent continuously updates its Q-values by:
α
Q(st , at ) ← rt+1 + λQ(st+1 , at+1 ) (2.1)
α
where x ← y ≡ x ← x + α(y − x) and λ is the discount parameter. Using
these evolving Q-values, the agent chooses the action with the highest Q(s, a) to
maximize its expected future rewards.
Particularly, the combination of DL and RL takes a step further in complex
system control. The traditional RL algorithm records the reward of (state, action)
through the table-based method, which will lead to complexity issues that the RL
is not designed for, namely memory complexity, computational complexity, and
sample complexity, as its use was inherently limited to low-dimensional problems
[31]. Specifically, in a large-scale and highly dynamic network, there are too
many (state, action) pairs. It is often impractical to maintain the Q-value for all
possible (state, action) pairs. Hence, it is common to use a parameterized function
Q(s, a; θ ) to approximate Q(s, a). Deep neural networks have powerful function
approximation and representation learning capabilities [32]. The DL algorithm
can automatically extract low-dimensional features from high-dimensional data.
Therefore, DL can effectively compress the network state space, as illustrated in
Fig. 2.6, thus enabling RL to solve large-scale network decision-making problems
that were previously found difficult in handling high-latitude states and motion
space.
Based on such reinforcement learning framework, the data flow in NetworkAI
is described as follow. The network monitor data and traffic information will be
collected by the upload link, the decision for each flow calculated in AI plane
send to SDN controller via northbound Interface. The SDN control then issue the
22 2 Intelligence-Driven Networking Architecture

Fig. 2.6 The compress process of mass states

rule through southbound interface. Thus, this data flow achieves RL agent through
interaction with the underlying network. Different applications just need craft the
reward signal to guide the agent toward good policy to meet their objectives.
In our architecture, we apply deep reinforcement learning to generate network
policy. Combining RL with DL leads to a general artificial intelligence solution for
solving complex network control problems. We believe that introducing DRL for
network decision making presents two main advantages.
First, the DRL algorithm is a black-box approach. The DRL agent only need to
have different network decision tasks and optimization goals in designing action
spaces and rewards without changing the mathematical model. In addition, because
an artificial neural network has the characteristic of expressing arbitrary nonlinear
mappings, the DRL agent can understand a nonlinear, complex, multi-dimensional
network control problem without the need of any simplifications. On the contrary,
traditional white-box approaches require assumptions and simplifications of the
underlying network aiming at building the equivalent mathematical model and
tailoring for a problem that has to be optimized.
Second, the DRL agent does not need to converge again when network state
changed [22]. Once the DRL agent trained, an approximate optimal solution can be
calculated in single step through matrix multiplication where the time complexity is
only approximately O(n2 ), where n is represented by number of network nodes. In
contrast, the heuristic algorithms need take many steps to converge to a new result,
where leads to high computational time cost. For example, the time complexity of
ant colony algorithm is O(n × (n − 1) × mt), where n is represented by number
of network nodes , m is number of ants , t is number of iterations. Therefore, DRL
offers a tremendous advantage for the real-time control of a dynamic network.
Above all, the NetworkAI achieves applying RL approach for the real-time
control of the network. The SDN, INT and traffic identification technologies are
used to implement the network state upload link and the decision download link
respectively with an aim to obtain a centralized view and control of a complex net-
work systems. In addition, DRL agent in AI plane can effectively solve complexity
network control problem without the need of any simplifications of real network
2.1 Network AI: An Intelligent Network Architecture for Self-Learning. . . 23

environment. Furthermore, that a near-optimal solution can quickly be calculated

once it is trained represents an important advantage for the real-time control of
the network. Thus, the NetworkAI facilitates an automated control of a large-scale
network.

2.1.3 Use Case

The main objective of this use case is to demonstrate that it is possible to model the
behavior of a network with the proposed NetworkAI architecture. In particular, we
present a simple example in the context of QoS routing, where the NetworkAI was
used to make intelligent decisions to select the best routing path aiming at satisfying
the QoS requirements.
The traditional Internet design is based on end-to-end arguments with an aim to
minimize the network supports. This type of architecture is perfectly suited for data
transmission where the primary requirement is reliability [10]. However, with the
proliferation of various applications (such as multimedia transmission application,
where timely delivery is preferred over reliability), the demands of each application
are different. Thus, the network should support QoS in a multi-application traffic
scenario [21]. However, the question of how to support end-to-end QoS is an on-
going problem.
QoS routing mainly involves path selection that meets the QoS requirements
arising from different service flows. It is a mechanism of routing based on QoS
requests of a data flow and the available network resources. The typical QoS
indicators for applications in the network are different, as demonstrated in Table 2.1
in which we list the QoS requirements for several applications.
The dynamic QoS routing can be seen as a Constrained Shortest Path (CSP)
problem, which is an NP-complete problem [33]. Although researchers from both
academia and industry have proposed many solutions to solve the QoS limitations
of the current networking technologies [19, 20, 33], many of these solutions
either failed or were not implemented because these approaches come with many
challenges. The tradition heuristic methods bring high computational time cost

Table 2.1 The QoS indicator for several applications

QoS
Application Delay Throughput Jitter Losses
√ √ √
Realtime multimedia
√ √ √ √
Augmented reality or virtual reality
√ √
VoIP
√ √ √ √
Scheduling in datacenters
√ √
Internet of vehicles
24 2 Intelligence-Driven Networking Architecture

Fig. 2.7 The experiment example

in network control, which is difficult to apply to real-time dynamic network

environment.
The NetworkAI paradigm can address many of the challenges posed by the QoS-
Routing problem. RL approach can calculate an near-heuristic solution in one step,
which can benefit for real-time control to large-scale network. And the closed-
loop architecture in NetworkAI provide an interaction framework which achieves
applying DRL in geo-distributed network. In our experiment, we applied the Deep
Q-learning algorithm in a central agent, aimed to select the best routing path to
minimize the network delay [21, 29, 34].
Methodology We simulated a network with 36 nodes and 50 full-duplex links,
with uniform link capacities and different transmission delay. One of these nodes
applied DQN and the others run short path algorithm. In addition, we devised ten
traffic intensity (TI) levels, ranging from 1 to 10, which represent the volumes of the
network traffic at different times. The traffic generated randomly in each node and
volumes is subject to poisson distribution.
As shown in Fig. 2.7, it is a part of the whole experiment topology. We applied
DQN to guide node i’s routing action. In our experiment, state is represented by the
link delay, node processing delay, which is up to 8620 states. Action is represented
by the tuple of node which univocally determine the paths for the source-destination
node pairs. As shown in Fig. 2.7, the action space for the node i to destination j is
(i, k, j ), (i, l, j ) and (i, m, j ). In addition, reward is represent by the delay from the
source to the destination node.
In our experiment, we employed a neural network that has five fully connected
hidden layers with a sigmoid activation function as well as a trained DQN on the
gym platform [35] and Keras [36]. In addition, we devised ten traffic intensity (TI)
2.1 Network AI: An Intelligent Network Architecture for Self-Learning. . . 25

50

45

40
Average Deliver Time

35

30

25

20

15

10

0
0 0.5 1 1.5 2
Traing Steps x 106

Fig. 2.8 The DRL learning process

levels, ranging from 1 to 10, which represent the volumes of the network traffic at
different times. The DRL agent was trained for 200K steps for each TI.
Experimental Results and Analysis The DRL learning process is demonstrated in
Fig. 2.8. This relevant outcome is that the DRL performance increases with training
steps and the DRL agent be convergence when the training step more than 200K.
The reason is that the DRL learning is a process that the result is close to a near-
optimal strategy by interacting with the environment. The second simulation results
are demonstrated in Fig. 2.9. It can be seen that the average transmission time of the
network increases with the increase of network traffic load. When the load level is
low, with the increase of network load, the increase of average transmission time is
stable. But when the network load continues to increase, the average transmission
time increases sharply, due to the fact that the network capacity is close to saturation.
In our experiment, the benchmark algorithm is the shortest path algorithm. When
the network load is low, there is no congestion in the network and the shortest path
is the optimal path. So, the benchmark result is better. With the increase of network
load, congestion occurs on the shortest path of the network, and the network agent
will have to choose non-congested links for transmission. Thus, in this situation, the
DRL performance performs much better than the benchmark.
26 2 Intelligence-Driven Networking Architecture

35
Shortest Path Routing

AI Routing
Average Deliver Time 30

25

20

15

10

5
1 2 3 4 5 6 7 8 9
Network Load

Fig. 2.9 The average delivery time over different network loads

2.1.4 Challenges and Discussions

In this subsection, we proposed a new network paradigm named NetworkAI. We

introduce this architecture to effectively solve the two main problems when apply
ML in network, how to run ML on distributed network and how to generate complex
strategies. Through the technologies of SDN, monitor technologies, and RL, the
intelligent agent can automatically generate network strategies for network control
and management. Moreover, through a use case, we have demonstrated that a fully
automated DRL agent can provide routing policy that tends to minimize the network
delay.
Compared with the traditional network architectures, we apply SDN and monitor
technologies to implement a completely centralized view and control for distributed
network systems and build a closed control loop for real-time upload of network
state info and download of actions.
At the same time, DRL is introduced to effectively generate control policy in
network systems. Benefit from the DL method with powerful function approxima-
tion and representation learning capabilities, the massive state space of the network
can be effectively compressed. In addition, due to the black-box feature of RL
approach, for different network decision tasks and optimization goals, only the
action space and reward have to be designed. Furthermore, the near-optimal solution
can be calculated in single step once the DRL trained. Compared with the traditional
heuristic algorithms, this feature presents a huge advantage for real-time network
applications.
2.1 Network AI: An Intelligent Network Architecture for Self-Learning. . . 27

The NetworkAI paradigm brings significant advantages to networking. However,

it also meets some challenges that need to be addressed further.

2.1.4.1 Communication Overhead

The communication overhead for retrieving and issuing data is a serious problem in
SDN architecture. While the convenience brought by centralized framework, it leads
to too much interaction between centralized controller and distributed forwarding
unit. The performance of NetworkAI will be degraded as a result of the rapid flow
table update to all forwarding unit. To address it, NetworkAI can borrow some
technologies from SDN. One possible solution is segment routing technology, which
implements the source routing and tunnel method to effectively reduce the flow table
update. Another way to alleviate this problem is to employ a cluster of controller to
handle larger flow tables [37].

2.1.4.2 Training Cost

RL approach provides flexible tools to address network control problems. Nonethe-

less, RL method involve a large amount of training cost, especially in network
scenario where exists mass applications and services. When new businesses or
services appears, the agent requires a significant level of training cost which weaken
the flexibility of NetworkAI. In this context, the NetworkAI paradigm requires
mechanisms to reduce the training cost to satisfy mass of new network applications.
A notable approach to address this problem is adding prior knowledge to accelerate
the training process, such as transfer learning, imitation learning [38]. In this sense,
these trick may reduce the training cost when new businesses or services appear and
essentially in taking a step further in performance of NetworkAI.

2.1.4.3 Testbeds

To evaluate the performance of new network designs and algorithms, testbeds are
more convincing for network experiments compared with simulators and emulation
platforms, because testbeds can incorporate real traffic and real network facilities
[39]. Building a complex experimental environment will be the most critical issue
for applying AI in a network. In particular, due to the fact that the NetworkAI
architecture is aimed at a complex, highly dynamic multi-application network
environment, it is difficult to obtain convincing experiments through the network
simulator. Therefore, in the immediate future, we plan to build a large-scale real
NetworkAI testbed to expand our experiments.
28 2 Intelligence-Driven Networking Architecture

2.2 Summary

In this chapter, we presents NetworkAI, an intelligent architecture for self-learning

control strategies in SDN networks. NetworkAI employs deep reinforcement learn-
ing and incorporates network monitoring technologies such as the in-band network
telemetry to dynamically generate control policies and produces a near optimal
decision.

References

1. O. N. Foundation, “Software-defined networking: The new norm for networks,” 2012.

2. C. Qiu, F. R. Yu, H. Yao, C. Jiang, F. Xu, and C. Zhao, “Blockchain-based software-defined
industrial internet of things: A dueling deep q-learning approach,” IEEE Internet of Things
Journal.
3. S. Sezer, S. Scott-Hayward, P. K. Chouhan, B. Fraser, D. Lake, J. Finnegan, N. Viljoen,
M. Miller, and N. Rao, “Are we ready for sdn? implementation challenges for software-defined
networks,” IEEE Communications Magazine, vol. 51, no. 7, pp. 36–43, 2013.
4. X. Lei, C. Jiang, N. He, H. Zhu, and A. Benslimane, “Trust-based collaborative privacy
management in online social networks,” IEEE Transactions on Information Forensics &
Security, vol. 14, no. 1, pp. 48–60, 2018.
5. H. E. Egilmez, S. T. Dane, K. T. Bagci, and A. M. Tekalp, “Openqos: An openflow controller
design for multimedia delivery with end-to-end quality of service over software-defined
networks,” in Signal & Information Processing Association Summit and Conference, pp. 1–
8, 2012.
6. S. Yussof and O. H. See, “The effect of ga parameters on the performance of ga-based qos
routing algorithm,” in International Symposium on Information Technology, pp. 1–7, 2008.
7. H. Yao, B. Zhang, P. Zhang, S. Wu, C. Jiang, and S. Guo, “Rdam: A reinforcement
learning based dynamic attribute matrix representation for virtual network embedding,” IEEE
Transactions on Emerging Topics in Computing.
8. D. Silver, A. Huang, C. J. Maddison, A. Guez, L. Sifre, G. V. D. Driessche, J. Schrittwieser,
I. Antonoglou, V. Panneershelvam, and M. Lanctot, “Mastering the game of go with deep
neural networks and tree search,” Nature, vol. 529, no. 7587, p. 484, 2016.
9. A. Mestres, A. Rodrigueznatal, J. Carner, P. Barletros, E. Alarcón, M. Solé, V. Muntés,
D. Meyer, S. Barkai, and M. J. Hibbett, “Knowledge-defined networking,” Acm Sigcomm
Computer Communication Review, vol. 47, no. 3, pp. 2–10, 2016.
10. H. Yao, C. Qiu, C. Fang, X. Chen, and F. R. Yu, “A novel framework of data-driven
networking,” IEEE Access, vol. 4, no. 99, pp. 9066–9072, 2017.
11. X. Lei, C. Jiang, N. He, Q. Yi, and J. Li, “Check in or not? a stochastic game for privacy
preserving in point-of-interest recommendation system,” IEEE Internet of Things Journal,
vol. PP, no. 99, pp. 1–1, 2018.
12. J. Jiang, V. Sekar, I. Stoica, and H. Zhang, “Unleashing the potential of data-driven network-
ing,” 2017.
13. C. Kim, A. Sivaraman, N. Katta, A. Bas, A. Dixit, L. J. Wobker, and B. Networks, “In-band
network telemetry via programmable dataplanes,” 2015.
14. D. Kreutz, F. M. V. Ramos, P. Esteves Verissimo, C. Esteve Rothenberg, S. Azodolmolky, and
S. Uhlig, “Software-defined networking: A comprehensive survey,” Proceedings of the IEEE,
vol. 103, no. 1, pp. 10–13, 2014.
15. “Pnda.” http://www.pnda.io/.
16. H. Mao, R. Netravali, and M. Alizadeh, “Neural adaptive video streaming with pensieve,” in
Conference of the ACM Special Interest Group on Data Communication, pp. 197–210, 2017.
References 29

17. C. Jiang, Y. Chen, Q. Wang, and K. J. R. Liu, “Data-driven auction mechanism design in iaas
cloud computing,” IEEE Transactions on Services Computing, vol. PP, no. 99, pp. 1–1, 2015.
18. C. Qiu, S. Cui, H. Yao, F. Xu, F. R. Yu, and C. Zhao, “A novel qos-enabled load scheduling
algorithm based on reinforcement learning in software-defined energy internet,” Future
Generation Computer Systems.
19. J. A. Boyan and M. L. Littman, “Packet routing in dynamically changing networks: a rein-
forcement learning approach,” in International Conference on Neural Information Processing
Systems, pp. 671–678, 1993.
20. S. C. Lin, I. F. Akyildiz, P. Wang, and M. Luo, “Qos-aware adaptive routing in multi-
layer hierarchical software defined networks: A reinforcement learning approach,” in IEEE
International Conference on Services Computing, pp. 25–33, 2016.
21. H. Zhang, B. Wang, C. Jiang, K. Long, A. Nallanathan, V. C. M. Leung, and H. V. Poor,
“Energy efficient dynamic resource optimization in noma system,” IEEE Transactions on
Wireless Communications, vol. PP, no. 99, pp. 1–1, 2018.
22. C. Fang, H. Yao, Z. Wang, P. Si, Y. Chen, X. Wang, and F. R. Yu, “Edge cache-based isp-cp
collaboration scheme for content delivery services,” IEEE Access, vol. 7, pp. 5277–5284, 2019.
23. N. Mckeown, T. Anderson, H. Balakrishnan, G. Parulkar, L. Peterson, J. Rexford, S. Shenker,
and J. Turner, “Openflow:enabling innovation in campus networks,” Acm Sigcomm Computer
Communication Review, vol. 38, no. 2, pp. 69–74, 2008.
24. P. Bosshart, M. Izzard, M. Izzard, M. Izzard, N. Mckeown, J. Rexford, T. Dan, T. Dan,
A. Vahdat, and G. Varghese, “P4: programming protocol-independent packet processors,” Acm
Sigcomm Computer Communication Review, vol. 44, no. 3, pp. 87–95, 2013.
25. “barefootnetworks.” https://www.barefootnetworks.com/.
26. A. W. Moore and D. Zuev, “Internet traffic classification using bayesian analysis techniques,”
Acm Sigmetrics Performance Evaluation Review, vol. 33, no. 1, pp. 50–60, 2005.
27. H. Zhang, L. Chen, B. Yi, K. Chen, Y. Geng, and Y. Geng, “Coda: Toward automatically
identifying and scheduling coflows in the dark,” in Conference on ACM SIGCOMM 2016
Conference, pp. 160–173, 2016.
28. T. T. T. Nguyen and G. Armitage, “A survey of techniques for internet traffic classification
using machine learning,” IEEE Communications Surveys & Tutorials, vol. 10, no. 4, pp. 56–
76, 2009.
29. M. Yue, C. Jiang, T. Q. S. Quek, H. Zhu, and R. Yong, “Social learning based inference for
crowdsensing in mobile social networks,” IEEE Transactions on Mobile Computing, vol. PP,
no. 99, pp. 1–1, 2017.
30. C. Li and C. Yang, “The research on traffic sign recognition based on deep learning,” in
International Symposium on Communications and Information Technologies, pp. 156–161,
2016.
31. K. Arulkumaran, M. P. Deisenroth, M. Brundage, and A. A. Bharath, “A brief survey of deep
reinforcement learning,” 2017.
32. J. Schmidhuber, “Deep learning in neural networks: An overview.,” Neural Networks the
Official Journal of the International Neural Network Society, vol. 61, p. 85, 2015.
33. M. Karakus and A. Durresi, “Quality of service (qos) in software defined networking (sdn): A
survey,” Journal of Network & Computer Applications, vol. 80, pp. 200–218, 2017.
34. S. Levine, S. Levine, S. Levine, and S. Levine, “Continuous deep q-learning with model-based
acceleration,” in International Conference on International Conference on Machine Learning,
pp. 2829–2838, 2016.
35. “gym.” https://gym.openai.com/.
36. “keras.” https://keras.io/.
37. L. Cui, F. R. Yu, and Q. Yan, “When big data meets software-defined networking: Sdn for big
data and big data for sdn,” vol. 30, pp. 58–65, 2016.
38. S. J. Pan and Q. Yang, “A survey on transfer learning,” IEEE Transactions on Knowledge &
Data Engineering, vol. 22, no. 10, pp. 1345–1359, 2010.
39. T. Huang, F. R. Yu, C. Zhang, J. Liu, J. Zhang, and J. Liu, “A survey on large-scale software
defined networking (sdn) testbeds: Approaches and challenges,” IEEE Communications
Surveys & Tutorials, vol. PP, no. 99, pp. 1–1, 2017.
Chapter 3
Intelligent Network Awareness

In the network, different applications produce various traffic types with diverse
features and service requirements. Therefore, in order to better manage and control
networking, the intelligent awareness of network traffic plays a significant role. Net-
work traffic information mainly includes service-level information (e.g., QoS/QoE),
anomaly traffic detection information, etc. In this chapter, we first present a multi-
level intrusion detection model framework named MSML to address these issues.
The MSML framework includes four modules: pure cluster extraction, pattern
discovery, fine-grained classification and model updating. Then, we propose a novel
IDS framework called HMLD to address these issues, which is an exquisitely
designed framework based on Hybrid Multi-Level Data Mining. In addition, we
propose a new model based on big data analysis, which can avoid the influence
brought by adjustment of network traffic distribution, increase detection accuracy
and reduce the false negative rate. Finally, we propose an end-to-end IoT traffic
classification method relying on deep learning aided capsule network for the sake
of forming an efficient classification mechanism that integrates feature extraction,
feature selection and classification model. Our proposed traffic classification method
beneficially eliminates the process of manually selecting traffic features, and is
particularly applicable to smart city scenarios.

3.1 Intrusion Detection System Based on Multi-Level

Semi-Supervised Machine Learning

With the rapid development of the Internet, the number of network invasions greatly
increases [1]. As a widely-used precautionary measure, intrusion detection has
become an important research topic. Machine learning (ML), which can address
many nonlinear problems well, has gradually become the mainstream in the field of

© Springer Nature Switzerland AG 2019 31

H. Yao et al., Developing Networks using Artificial Intelligence,
Wireless Networks, https://doi.org/10.1007/978-3-030-15028-0_3
32 3 Intelligent Network Awareness

intrusion detection system. Many existing models based on ML employ supervised

learning algorithms to train an intrusion detection classifier using a set of labeled
training samples, then use this classifier to classify unlabeled test data.
There are two main problems in network traffic, which affect the robustness of the
ML model. Firstly, network traffic has a severe class-imbalance problem. It means
that some categories have much more samples than others. This problem is also
called ‘elephant traffic’ and ‘mouse traffic’ problem. The generated ML model will
suffer because the model will be much more suitable for ‘elephant traffic’ rather
than ‘mouse traffic’. Secondly, training data and test data can have a non-identical
distribution problem. It means that our training data and test data are generated
by two different probabilistic distributions [16, 31]. The independent identical
distribution is an extremely important premise of statistical machine learning. Non-
identical distribution problem can cause a decrease in accuracy rate. Unfortunately,
the distribution of network traffic is not static because the uncertainty of users’
behaviors leads to the variable distribution of network traffic. The high cost of expert
system determines that it is not possible to mark a large amount of network traffic
in real-time [39]. We usually train the model using historical labeled data.
In this section, we propose a novel intrusion detection system based on a
multi-level semi-supervised machine learning framework (MSML) to tackle with
aforementioned problems. The main contributions of this part are described as
follows:
• In order to alleviate the class imbalance problem, we propose a concept of
pure cluster, and propose a cluster-based under-sampling methods, which is a
hierarchical, semi-supervised k-means clustering algorithm. There exists a large
number of samples in pure cluster. Furthermore, these pure cluster samples can
be predicted accurately using our proposed method.
• For the non-identical distribution problem, a method is proposed to distinguish
known and unknown pattern samples in the test set. We can obtain a guaranteed
high recognition accuracy for known pattern samples. Also, the unknown pattern
samples could be fine-grained classified.
In this part, the MSML framework is experimented and evaluated on the KDD-
CUP99 dataset. The results show the proposed framework significantly outperforms
our baseline method and many other existing models in the respect of overall
accuracy, F1 score and unknown pattern discovery ability. The results also show
that when the non-identical distribution problem do not occur, our framework is
still suitable and well-performed.

3.1.1 Proposed Scheme (MSML)

In this subsection, we introduce our proposed MSML framework, as shown in

Fig. 3.1. The data generator process for MSML is also shown in Fig. 3.1.
3.1 Intrusion Detection System Based on Multi-Level Semi-Supervised. . . 33

network flow generator Training Training labeled

P(x) unlabeled data data

Training data
Test data

Data preprocess
Data Generator Process

Pure Cluster Extraction

None?
No
Yes Model
finished
Pattern Discovery
No Updating
supervised known unsupervised
New?
classfier compensation clustering

Yes Fine-grained Classification

unsupervised mannal
classfier inspection

finished
MSML

Fig. 3.1 The MSML framework

The aim for data generator process is to generate the required training set and test
set for the MSML framework. Due to the semi-supervised property of the MSML,
the training set consists of labeled samples and unlabeled samples. The training
labeled data was labeled in the past, reflecting the distribution of historical known
network traffic. The training unlabeled samples and test samples are all generated
by the network traffic generator, which reflects the distribution of current network
traffic. The data preprocessing module is devised to do something necessary before
training model such as normalization and data cleaning.
The MSML consists of four modules including pure cluster extraction, pattern
discovery, fine-grained classification and model updating. The pure cluster extrac-
tion module aims to find large and pure clusters. In the pure cluster module, this
section defines an important concept of “pure cluster pattern” and proposes a hierar-
chical semi-supervised k-means algorithm (HSK-means), aiming at finding out all
the pure clusters. In the pattern discovery module, this section defined the “unknown
pattern” and applied cluster based method to find those unknown patterns. The fine-
grained classification module achieves fine-grained classification for those unknown
pattern samples. The model updating module provides a mechanism for retraining.
For any test sample, once labeled by one module, will not go on any more; and
all test samples will be labeled in pure cluster extraction module, pattern discovery
module and fine-grained classification module.
34 3 Intelligent Network Awareness

3.1.1.1 Pure Cluster Extraction (PCE)

The pure cluster can be defined as a cluster where almost all samples have the same
category. The potential samples located in this pure cluster can be considered to be
the same category as the other samples in this cluster [33].
Given a training labeled set, D l = {l1 , l2 , · · · , lN }, which consists of N labeled
samples and a unlabeled training set, D u = {u1 , u2 , · · · , uM }, which consists of
M unlabeled samples. The labeled and unlabeled training samples are merged with
an aim to form our training set D = D l ∪ D u . The training set are partitioned
into K clusters {C1 , C2 , . . . , CK } (K ≤ |D|) using the K-means clustering method.
If the training labeled samples and the training unlabeled samples are identically
distributed, we assume that a large cluster Ci which contains a large number of
samples will meet the following formula, by the principle of the central limit
theorem.

|Cil | N
≈ , (3.1)
|Ci | N +M

where Cil is the k-th cluster in the labeled samples. We can conclude that the cluster
is not pure, if the left hand side of Eq. (3.1) is significantly smaller than the right
hand side. The reason is that the labeled samples only characterize a part of the
whole cluster. Consequently, we mark a cluster “pure” cluster when all the labeled
samples of this cluster belong to a same category and the cluster meets the following
formula:

|Cil | N
≥η∗ , (3.2)
|Ci | N +M
|Ci | ≥ MinP C, (3.3)

where MinP C is a parameter reflecting the minimum number of samples of a

cluster which is required to be spilt using a cluster method recursively. η is a real
number and cannot exceed 1. However, the value of η cannot set to be too small. We
proposed a hierarchical semi-supervised K-means (HSK-means) based on the above
mentioned definition of pure cluster. The pseudo-codes of training phases of HSK-
means algorithm are shown in Algorithm 1. We define a new parameter ArsP C,
reflecting the average number of samples of clusters when HSK-means employ the
clustering in all the samples of a parent cluster. The formula (3.4) shows the method
of calculating the number of child clusters. The key point is that if a cluster do not
meet formula (3.2), then we will perform the K-cluster algorithm on this cluster
recursively unless this cluster do not meet formula (3.3), as is shown in the 10th line
in Algorithm 1. The pseudo-codes of training phases of HSK-means algorithm are
shown in Algorithm 2.

|C|
K= . (3.4)
ArsP C
3.1 Intrusion Detection System Based on Multi-Level Semi-Supervised. . . 35

Algorithm 1 MSML-PCE HSK-means training

Require: training set D
Ensure: new training set D ∗
1: function TREECLUSTER(D)
2: Calculate the number of clusters using formula (3.4)
3: Perform clustering on D to obtain clusters {C1 , C2 , . . . CK }
4: D.childern ← {C1 , C2 , . . . CK }
5: for i = 1 → K do
6: if the cluster Ci is a pure cluster then
7: Label all samples in Ci
8: Update D ∗ with D − Ci
9: else if |Ci | ≥ ArsP C then
10: T reeCluster(Ci )
11: end if
12: end for
13: end function

Algorithm 2 MSML-PCE HSK-means test

Require: test set T
Ensure: new test set T ∗ , labeled test set Tf inished
1: for ∀si ∈ T do
2: index ← D
3: while index has children do
4: Find the nearest children Cj from si
5: index ← Cj
6: if index has a label C then
7: Label si with C
8: Put si into T ef inished
9: break
10: end if
11: end while
12: if si has not a label then
13: Put si into T ∗
14: end if
15: end for

The time complexity of the HSK-means algorithm is O(ntk), where n is the number
of all the samples in the training set, t is the number of iterations with the general
k-means, k is the number of all the clusters including those generated by their
parent clusters and those generated by the whole data. The HSK-means algorithm is
faster than general k-means when both of clusters are identical. In the HSK-means
algorithm, some of clusters is generated by their parent clusters, the number samples
of which is obviously much less than the whole data [36].
For the training set, any sample in pure clusters can be extracted from the training
set. We only preserve those non-pure cluster samples for the training set of the next
module. This is actually a under-sampling method based on cluster. For those pure
clusters, the sampling rate is zero. For those non-pure clusters, the sampling rate is
one. For the test set, all samples in the pure clusters are labeled while all samples in
36 3 Intelligent Network Awareness

those non-pure cluster are not labeled and are preserved for the test set of the next
module.
If all pure clusters cover overmuch samples, the module will leave little
remaining training data to next module, thereby falling into overfitting. Hence,
the appropriate values of two parameters ArsP C and MinP C are important for
the whole MSML framework. It is necessary to adjust these values of the two
parameters.

3.1.1.2 Pattern Discovery (PD)

Pattern is an abstract concept. In this chapter, it refers to a kind of data distribution

in the feature space. Pattern can be classified in different ways [40]. For example,
pattern can be divided into known pattern and unknown pattern. The known patterns
refer to all patterns that exist in the training set, while the unknown patterns refer
to patterns that do not exist in the training set. For another example, pattern can
also be classified into intrusion pattern and normal pattern. The intrusion patterns
refer to patterns of labeled training intrusion samples, while the normal patterns
refer to patterns of labeled training normal samples. In a word, pattern is an
abstract description about data distribution that we wish to recognize. What we
wish to recognize determines the definition of specified pattern. Then we can apply
heuristic and machine learning methods to recognize it. For example, the pure
cluster introduced in Sect. 3.1.1 are belonged to a kind of pattern. Then we use
our proposed hierarchal semi-supervised K-means algorithms to obtain the required
pure clusters.
In this chapter, the main criteria that determines a sample to be known or
unknown is the pattern of this sample rather than the sample category. Our
hypothesis about the real distribution of network traffic is that the majority of known
traffic samples come from known patterns while the majority of unknown traffic
samples come from unknown patterns. That means, the probability that a known
traffic sample is an unknown pattern sample is very close to zero. It is so distinctive
to the work in [7, 9] where the criteria is traffic category. From the point of view of
patterns, [6, 8, 9] indicated an assumption that unknown traffic was unknown pattern
and known traffic was known pattern.
This module has three sub-modules, clustering, known compensation and super-
vised classifier. The pseudo-codes of training and test phases of the pattern recovery
module are shown in Algorithm 3. All training samples which are not located in
pure clusters form this module’s training set. Similarly, all test samples which are
not labeled in previous module form this module’s test set.
We applied a semi-supervised algorithm named K-means in the process of
clustering. K-means employ a clustering in the whole training set T rpd . We define
a new parameter AsPD to denote the average size of clusters . Obviously, the
number of clusters for the clustering can be expressed by K = |T rpd | / AsP D.
If the training labeled samples and the training unlabeled samples are identically
distributed, a large cluster, which contains a large number of samples, will meet
3.1 Intrusion Detection System Based on Multi-Level Semi-Supervised. . . 37

Eq. (3.1) by the central limit theorem. However, if a cluster does not meet Eq. (3.1)
but it meets the following formula, we regard the cluster as a unknown pattern and
label all the samples in the cluster as “new”.

|Cil | N
<μ∗ (3.5)
|Ci | M
|Ci | ≥ AsP D (3.6)

where μ is a real number. Its value is suggested to a small number.

The second step is called known compensation. With the decrease of the value
of AsPD, both TP (true positive rate) and FP (false positive rate) increase. The
reason is that some known patterns are considered as unknown patterns. This
can significantly reduce the precision of the unknown patterns and can increase
the complexity of the internal structure of the unknown patterns. In order to
maximize the increase of TP while minimizing the increase of FP, we make some
compensation. The idea is to use other supervised learning algorithms to train the
labeled data and calculate the confidence of each unknown patterns. Those patterns
whose confidence are large enough will not be considered as unknown patterns
any more. We define a confidence minimum threshold. In this chapter, we use the
softmax regression algorithm to calculate this confidence. Considering the softmax
regression is sensible to unbalanced data, this section presents a cluster-based under-
sampling technique applying a under-sampling rate function. The larger the cluster
is, the value of the under-sampling rate function is.
The third step is “supervised classifier”. After the second step, the training set
is composed of one normal category, N intrusion categories, and one unknown
category. The normal and intrusion categories are all regarded as known patterns.
The unknown category represents all the unknown patterns. In this chapter, we use a
random forest algorithm with an aim to build a supervised classifier. After this step,
all the test samples are labeled. We successfully separate all the test samples into
known patterns and unknown patterns. If unknown patterns are essential to perform
a fine-grained classification, then next model will work.
The time complexity of Algorithm 3 depends on the time complexities of its three
sub algorithms, including K-means, logistic regression and random forest. Its time
complexity equals to the sum of the time complexities of the three sub algorithms.

3.1.1.3 Fine-Grained Classification (FC)

After the processing of pattern discovery module, all the test samples are labeled.
However, some samples are labeled as “new”, which is neither a normal category nor
an intrusion category, but a new category. Expert inspection is used to achieve fine-
grained classification. We have high confidence to classify these samples correctly
with low artificial cost because we have separated a few of unknown patterns from
a group of complicated patterns. Algorithm 4 is the pseudo-code of fine-grained
38 3 Intelligent Network Awareness

Algorithm 3 MSML-PD
Require: training set D ∗ , test set T ∗ , labeled test set Tf inished
Ensure: labeled test set Tf inished , unknown test set T
1: T ← {}, Ds∗ ← {}, Dn ← {}
2: Perform clustering on D ∗ to obtain clusters {C1 , C2 , . . . CK }
3: for i = 1 → K do
4: if Ci is an unknown
 pattern then
5: Dn ← Dn Ci
6: end if
7: Generate Cis ⊆ Ci
8: Put q into Ds∗ , ∀q ∈ Cis
9: end for
10: Choose training labeled samples Dl∗ from Ds∗
11: Train a softmax classifier using Dl∗
12: for all Ci ∈ Dn do
13: if ∃ category c & minimum probability of c ≥ α then
14: Dn ← Dn − Ci
15: end if
16: end for
17: Choose training labeled samples Dl from D ∗
18: Combine Dl with Dn to train a supervised classifier f
19: for ∀s ∈ T ∗ do
20: if s is classified as new by classifier f then
21: Put s into T
22: elsePut s into Tf inished
23: end if
24: end for

Algorithm 4 MSML-FC
Require: labeled test set Tf inished , unknown test set T
Ensure: labeled test set Tf inished which includes some additional samples
1: Perform clustering on T to obtain clusters {C1 , C2 , . . . CK }
2: for i = 1 → K do
3: count ← 0
4: while count < MaxF C do
5: Randomly select A samples from Ci
6: Expert inspection
7: if all the A samples has the same ground-truth category C then
8: Label all samples in Ci with C;
9: break;
10: end if
11: end while
12: if none sample in Ci is labeled then
13: Label all samples in Ci with suspious
14: end if
15: Put all samples in Ci into Tf inished
16: end for
3.1 Intrusion Detection System Based on Multi-Level Semi-Supervised. . . 39

classification module. we perform k-means clustering on these unknown pattern

samples first. For each cluster, we randomly select several samples (e.g., three
samples) for manual inspection. If all the selected samples have the same ground-
truth category, then we achieve fine-grained classification, as is shown in 5–9 lines
in Algorithm 3. The parameter MaxF C indicates the number of allowed attempts.
The time complexity of Algorithm 4 depends on the time complexity of the K-
means and the velocity of expert inspection process.

3.1.1.4 Model Updating

In this section we continue to discuss how we can update our model. When the
amount of “new” samples is relatively large, it is possible to train a supervise model
based on the samples of these unknown patterns. In this manner, a consecutive
sample of the “new” can be identified as a specific class directly by this model. Only
the current distribution of network traffic generator does not change, it is effective
to do so.
When the distribution varies due to going through a long period of time,
introducing a feedback mechanism is necessary. If a new cluster is pure and have
enough samples and the cluster does not overlap with other prepared pure clusters in
feature space, then it is time to update the pure cluster extraction module. Otherwise
we should update the pattern discovery module. In doing so, the model can be
always able to adapt to the new traffic distribution.

3.1.1.5 The Hyper-Parameters

The main hyper-parameters of MSML are listed as shown in Table 3.1.

Table 3.1 Main hyper-parameters

Name Module Definition
ArsPC MSML-PCE The average number of samples of clusters
MinPC MSML-PCE The minimum number of samples of a
cluster which is required to be spilt using a
cluster method recursively
ArsPD MSML-PD The average number of samples of clusters
MaxFC MSML-FC The number of attempts to conduct
inspection allowed
40 3 Intelligent Network Awareness

3.1.2 Evaluation
3.1.2.1 Dataset

We choose the KDDCUP99 dataset to evaluate the MSML framework [43]. The
KDDCUP99 dataset contains four intrusion categories and one normal category.
The four intrusion categories are DOS, R2L, U2R and Probe, respectively. Each
of intrusion categories contains several of subcategories. The KDDCUP99 training
dataset contains 5 categories and 22 subcategories. The KDDCUP99 test dataset
contains 17 subcategories of 4 categories which do not appear in the training dataset.
The reason why we choose KDDCup99 as our dataset consists of two aspects. The
one is that the source of dataset we can get is limited, the other one is that there
are a large number of research works using this dataset can be compared with our
proposed method.
Inconsistent Dataset
In order to evaluate the performance of our MSML framework on non-identical
distribution dataset, we construct a dataset named inconsistent dataset. The training
dataset is composed of two parts including labeled samples and unlabeled samples.
We choose a fraction of the KDDCUP99 training dataset as labeled samples due to
the fact that using all of the KDDCUP99 training dataset is time-consuming. We
randomly select 20% of the KDDCUP99 test dataset as unlabeled samples due to
the fact that the training unlabeled samples and test samples need to be identically
distributed according to our MSML framework. Table 3.1 shows the composition of
the compositive training set of inconsistent dataset (Table 3.2).

Consistent Dataset
In order to evaulate the performance of our MSML framework on identical
distribution dataset, we construct a dataset named consistent dataset. We divide
the KDDCUP99 training dataset into three subsets according to the rate of 20%,
20% and 60%, respectively. These three subsets represent training labeled samples,
training unlabeled samples and test samples, respectively.

Table 3.2 Composition of Category Number

training set
Normal 3242
DOS 1026
Probe 7829
U2R 52
R2L 563
Unlabeled 60,020
3.1 Intrusion Detection System Based on Multi-Level Semi-Supervised. . . 41

3.1.2.2 Data Pre-process

The KDDCUP99 dataset contains 41 features including nine discrete features and
32 consecutive features. We adopt one-hot and numerical-order methods to deal
with discrete features. We employ one-hot when using ANN and K-means due to
the fact that numerical-order can bring nonexistent sequence influence. We adopt
numeric-order when using other methods because one-hot can greatly increase the
data sparsity.
In this chapter, 32 consecutive features are normalized. However, the
values of some consecutive features (“duration”, “src_bytes”, “num_root”,
“num_compromised”, “_bytes”) show unusual value distribution. For these unusual
features, the majority of the values are much smaller than the maximum value,
which means 0–1 normalization will make majority of values close to zero. To
avoid this situation, we adopt logarithmic normalization processing. Logarithmic
normalization does not change the order of the values, but it can significantly reduce
the effect of abnormal maximum value.

3.1.2.3 Evaluation Criteria

The TP, TN, FP, FN[10] are usually used to evaluate the performance of machine
learning model, which can be described by a confusion matrix shown in Table 3.3.
The Precision, Recall, F1_score and Accuracy [10] are also defined to evaluate
the model performance.

TP
P = P recision = (3.7)
T P + FP

TP
R = Recall = (3.8)
T P + FN

2 × P recision × Recall
F 1_score = (3.9)
P recision + Recall

TP +TN
Accurary = (3.10)
T P + T N + FN + FP

In addition, we define several of evaluation indexes for MSML framework

including Capture rate, Coverage rate and Coverage capture rate. After pattern

Table 3.3 Confusion matrix Predicted

Actual Positive Negative
Positive TP FN
Negative FP TN
42 3 Intelligent Network Awareness

Table 3.4 Self-defined indexes

Index name Capture rate Coverage rate Coverage capture rate
Index value B/M b/M b/B

Table 3.5 Confusion Predicted

matrix(baseline, inconsistent
Actual Normal DoS R2L U2R Probe
dataset)
Normal 60,255 743 23 13 259
DoS 6203 223,364 8 1 277
R2L 15,627 0 539 17 6
U2R 134 0 8 30 56
Probe 620 154 2 0 3390

discovery module, Capture rate is the proportion of all test samples which are
labeled. Coverage rate is the proportion of all test samples which are correctly
labeled. Coverage capture rate is the proportion of test labeled samples which are
correctly labeled. We suppose the test set has M samples. After performing the
operation of pattern discovery module, we classify B samples, therein, we correctly
classify b samples. Apparently, the remaining M − B samples are labeled “new”,
waiting for being fine-grained labeled. The calculation ways of three indexes are
shown in Table 3.4.

3.1.2.4 Baseline Model

In order to validate the effectiveness of the MSML framework, we adopt traditional

1+N model as our compared baseline model. Our experiment based on the baseline
model is carried out on both inconsistent dataset and consistent dataset.
Inconsistent Dataset Comparison
We adopt the labeled samples from the training set of inconsistent dataset to train
the baseline model, and evaluate the baseline model on the test set of inconsistent
dataset. We take use of several supervised learning algorithms, including Naive
Bayes, BP neural network, random forests, support vector machines and so on.
Experimental results indicate that random forest achieves the highest overall
accuracy by 92.46%. Table 3.5 shows the confusion matrix on the test set.
Furthermore, through the deep analysis of Fig. 3.2, we can conclude that the
overall accuracy of the known traffic samples can reach 97.94%. However, the
accuracy of unknown traffic samples is only 6.87%. The unknown traffic samples
cause a decrease in not only overall accuracy, but also recall rate of all categories,
as shown in Fig. 3.2.
Consistent Dataset Comparison
An experiment with the similar method as on the inconsistent dataset is conducted
on the consistent dataset. We adopt the labeled samples of the consistent training
3.1 Intrusion Detection System Based on Multi-Level Semi-Supervised. . . 43

Fig. 3.2 Recall rate comparison of each category in baseline model

dataset to train our model. We employ this trained model to evaluate on two test
set. One is the whole test set of the consistent dataset. We obtain 99.92% overall
accuracy. 99.92%, which is so close to 100%, representing the high recognition
capability of known pattern samples. The other is the unknown traffic portion of
test set of the inconsistent dataset. We obtain 97.9% accuracy, worse than the value
of 99.92%. We attribute the difference to non-identical distribution. Some known
traffic samples in the test set of inconsistent dataset actually belong to unknown
patterns thereby deteriorating the overall accuracy.

3.1.2.5 MSML

In addition to the overall accuracy, this part claims that the capture rate and coverage
capture rate are also significant. We make the index of coverage capture rate have
a high value in order to reduce the influence of error classification on intrusion
detection. The two indexes of coverage capture rate and capture rate are often
contradictory, due to the fact that the model is apt to consider a portion of known
pattern samples as unknown pattern samples. In this way, it can lead to a decrease
for the index of capture rate and an increase for the index of coverage capture rate.
44 3 Intelligent Network Awareness

Meanwhile, it can greatly increase the complexity of the structure of internal feature
space for unknown pattern samples and increase the burden on the subsequent fine-
grained classification module.
In this part, there are several parameters which may have an important impact
on the above two indexes of coverage capture rate and capture rate. For example,
the average cluster size of pure cluster extraction module denoted by ArsP C, the
lower limit cluster size of pure cluster extraction module denoted by MinP C, and
average cluster size of pattern discovery module denoted by AsP D, these three
parameters are of great concern to us. We set different values to these parameters in
our experiments on both inconsistent dataset and consistent dataset.
Inconsistent Dataset Comparison
We conduct some experiments on the inconsistent dataset. The parameter AsP D is
set to 20, 50, and 100, respectively. Simulation experiments show that the parameter
values have little effect on the results. In our work, we set AsP D to value of 100.
Figures 3.3 and 3.4 respectively show the trend of coverage capture rate and capture
rate with the change of ArsP C and MinP C. In these figures, each circle represents
a value of coverage capture rate and capture rate, the more red the circle is, the
bigger the circle is, the higher the circle value becomes.
Figure 3.3 reflects the relationship between coverage capture rate and ArsP C,
MinP C. From Fig. 3.3, we can observe that the coverage capture rate is generally

Fig. 3.3 MSML: coverage capture rate

3.1 Intrusion Detection System Based on Multi-Level Semi-Supervised. . . 45

0.90
Low limit size of pure cluster of MSML-PCE MinPC

2000 0.85

1500
0.80
1000

800
0.75
500

300 0.70
200

150 0.65

100
0.60
50

20
0.55

0.50
50 100 150 300 500 700 900 1200 1500 2000
Average size of clusters of MSML-PCE ArsPC

Fig. 3.4 MSML: capture rate

97% at least. In addition, with the increase of ArsP C and MinP C, the coverage
capture rate has a tendency of sightly increase first then significantly decrease.
Figure 3.4 reflects the relationship between capture rate and ArsP C and
MinP C. From Fig. 3.4, we can observe that the capture rate is relatively low when
ArsP C and MinP C have lower values. In addition, the overall trend of capture
rate is increasing with the increase of ArsP C and MinP C. Furthermore, there is
a certain randomness when the values of ArsP C and MinP C are large, therefore,
when the values of ArsP C and MinP C are bigger than a predefined threshold, it
is not necessary to continue to increase their values.
Considering both coverage capture rate and capture rate, it is proper to choose
a lower ArsP C and a larger MinP C. In this chapter, the parameter ArsP C is set
to 100 and the parameter MinP C is set to 1500. The accuracy of MSML can reach
96.6%, which has greatly improved compared with the baseline model, which can
reach 92.5%. It can be seen in Table 3.6.
In the fine-grained classification module, 150 samples, whose proportion is less
than 0.05%, are conducted expert inspection. The result of expert inspection is then
applied to all the unknown pattern samples, whose proportion is about 12%. The
accuracy of unknown pattern samples is 76.7%.
With respect to the recall rate, as is shown in Fig. 3.5, the elephant traffic DOS
has improved to some extent, while the common traffic has a great improvement.
For the mouse traffic, U2R and R2L have particularly notable improvement. The
46 3 Intelligent Network Awareness

Table 3.6 Comparison

Method Normal DoS Probe R2L U2R Accuracy
SVM+ELM+MK[13] 98.1 99.5 87.2 21.93 31.39 95.79
NFC[11] 98.2 99.5 84.1 14.1 31.5 N/A
SVM+BIRCH[35] 99.3 99.5 97.5 19.7 28.8 95.7
MOGFIDS[2] 98.4 97.2 88.6 15.8 11.0 93.2
Association rules[12] 99.5 96.6 74.8 3.8 1.2 92.4
Baseline 99.4 97.2 81.4 3.3 13.2 92.5
MSML 86.2 99.8 98.4 90.6 72.5 96.6
Bold values indicate the maximum value of each column

Fig. 3.5 MSML: recall

recall rate of U2R and R2L have greatly increased from 13.2% to 72.5%, and 3.3%
to 90.6%, respectively. With respect to the precision, as is shown in Fig. 3.6, the
precision of DOS remains 99.9%, and the precision of Normal, U2R, and R2L have
improved at different degree. With respect to the F1 score, as is shown in Fig. 3.7,
the F1 score of all categories improves. The F1 score of Normal, Dos, Probe, R2L
and U2R have increased from 0.885 to 0.912, from 0.985 to 0.999, from 0.832 to
0.934, from 0.064 to 0.754 and from 0.208 to 0.743, respectively.
However, we must note that the recall rate of normal and the precision rate of
R2L have a descending trend. This situation should be further investigated. Another
experiment is conducted. In the model updating module, we find a suspicious
cluster, where mixed snmpgetattack and snmpguess of R2L with Normal samples.
The number of R2L samples and the number of normal samples is about 53 to 47.
3.1 Intrusion Detection System Based on Multi-Level Semi-Supervised. . . 47

Fig. 3.6 MSML: precision

Fig. 3.7 MSML: F1-score

48 3 Intelligent Network Awareness

Further study[13] find that, in this cluster, samples of R2L and Normal in the feature
space are highly analogous so it is difficult to expect them distinguished. To confirm
this conjecture, we make all the samples of the suspicious cluster randomly divided
into a training set and a test set. We train a supervised classifier using the training
set, and evaluate on the test set. Experimental result show that the accuracy of the
test set is no higher than 53%. It is illustrated that samples of R2L and normal really
cannot be separated in this kind of cluster. We take the principle of priority of the
invasion. Hence, all the samples in this cluster are sentenced to R2L. This is different
from the baseline experiment, where all of samples are adjudged to Normal. It is the
reason why the recall rate of normal and the precision rate of R2L decrease.
Figures 3.8 and 3.9 show the relationship between categories and patterns. Figure
3.8 shows the ratio of known traffic to unknown traffic in the whole test dataset, in
the unknown pattern samples, and in the known pattern samples respectively. It can
be seen that the unknown traffic ratio of unknown pattern samples to the whole
test dataset has a great increment. We can observe that there are 89% of unknown
traffic which is considered as unknown pattern. Meanwhile, 5% of known traffic is
considered as unknown pattern. Figure 3.9 shows the ratio of known pattern samples
to unknown pattern samples in the whole test set, in the unknown traffic, and in the
known traffic, respectively.
We conduct two experiments with an aim to verify that the HSK-means algorithm
in the pure cluster extraction module is important and indispensible in our MSML
framework. One experiment begins with the pattern discovery module lacking of the
pure cluster extraction module. The result is shown as Fig. 3.10. From Fig. 3.10 we

Fig. 3.8 Pattern in traffic

3.1 Intrusion Detection System Based on Multi-Level Semi-Supervised. . . 49

Fig. 3.9 Traffic in pattern

can see that the capture rate decline and coverage capture rate rise with the increase
of the number of the cluster in pattern discovery module. However, when the
coverage capture reaches more than 99% as we anticipate, the capture rate is about
78%, which is much smaller than the counterpart which can reach about 88% of the
MSML framework. The other experiment replaces the HSK-means algorithm with a
common under-sampling method. The result, as is shown in Fig. 3.10, demonstrated
that the coverage rate cannot reach 96.7% in any case. We can conclude that the pure
cluster extraction is important and indispensible to our MSML framework based on
the results of these two experiments.
The performance comparison of MSML and other models is illustrated in
Table 3.6. MOGFIDS and baseline models are the most common “1+N” supervised
learning models. Both of them have a good detection rate in Normal and DOS, but
the detection rate of U2R and R2L is very bad. Furthermore, the overall accuracy is
also at a low level. Association rules[12] is a pure unsupervised learning algorithm
combining heuristic rules that can hardly be identified by the rat categories. Both of
[13] and [35] apply ways to identify unknown samples. Their detection rate of DOS
is particularly high, and the detection rate of other categories is also at a high level.
For our MSML, the detection rate of DOS, Probe, U2R and R2L and the overall
accuracy are the highest. We has been analyzed that it is the defects of KDDCUP99
test set that lead to a decrease for the detection rate of Normal in MSML. Therefore,
we can make the conclusion that MSML-IDS has a strong robustness.
50 3 Intelligent Network Awareness

Fig. 3.10 The importance of HSK-means

Consistent Dataset
The framework of MSML is also employed in the consistent dataset. In this dataset,
whatever the values of ArsP C, MinP C, and AsP D are, we obtain a capture rate
of 99.95%. The rate of unknown pattern samples is less than 1/30,000, which can be
neglected. The experiment indicates that MSML can also be applied to distributed
consistent dataset, due to the fact that MSML does not easily classify known pattern
samples into unknown patterns. For the difference between training dataset and test
dataset, MSML shows a good adaptability.

3.2 Intrusion Detection Based on Hybrid Multi-Level Data

Mining

The development of mobile network is a double-edged sword, which brings both

convenience and network security problem to us. Therefore research about network
security makes great sense [3]. Traditional network security technologies such as
firewall, data encryption, and user authentication system are all passive defense
techniques. These methods have a good performance for known attacks, but not
3.2 Intrusion Detection Based on Hybrid Multi-Level Data Mining 51

for unknown attacks. Different from traditional technologies, Intrusion Detection

System (IDS) [15], which is firstly proposed by Denning et al. in 1986 [17], is an
innovative and proactive network security technology, meaning that it can detect
both known and unknown attacks. The basic architecture of IDS consists of three
parts: data acquisition part, intrusion detection part and response process part. Data
acquisition part of IDS is used to collect data from internet. The collected data,
which consists of normal data and many different types of attacks, is then send to
intrusion detection part. Intrusion detection part needs to pick out attack data from
normal data and identify what type of the attack is. Response process part receives
the detected attacks and processes them according to their types. As the core of IDS,
intrusion detection part has become a hot topic in research in recent years.
The widely use of network produces massive data which is a valuable resource.
From these data, we can extract much insightful information through data analysis
techniques [4, 18, 25, 28]. We name the data collected by data acquisition part
‘intrusion data’. Intrusion data has some features that will affect the performance
of IDS. We summary these features as follows:
• Multicomponent: The intrusion data is multicomponent because there exists
many types of attacks in it. Hence intrusion detection is not a binary-classification
problem, but a multi-classification problem, which is more complicated.
• Data imbalance: The number of attacks in different categories vary greatly
from each other. It means that the intrusion data has a severe problem of data
imbalance. Some attacks are difficult to be detected due to its sparseness.
• Time-varying: The distribution of data is time-varying. The variety between
training data and detecting data will affect the performance of IDS.
• Unknown attacks: Some new attacks may appear as time goes by. These unknown
attacks are difficult to be detected.
In recent years, researchers begin to use the artificial intelligence (AI) technology
to deal with the problems caused by the above mentioned features. Machine learning
(ML), whose main idea is building model to mine information from data, is a
core subfield of AI including many algorithms [49]. It can be roughly divided into
supervised learning algorithms and unsupervised learning algorithms according to
whether it has a training phase or not. Supervised learning is the most widely-
used technique in machine learning such as Support Vector Machine (SVM) [19],
Artificial Neural Network (ANN) [20], Decision Tree (DT) [21], Random Forest
(RF) [22] and so on. Unsupervised learning mainly refers to clustering algorithms
[23] such as K-Means[24], DBSCAN[26], Affinity Propagation (AP)[27] and so
on. Every machine learning algorithm has its own advantages and disadvantages.
Combining two or more algorithms together and taking full use of their advantages
will increase the performance of IDS. Therefore, how to combine these algorithms
scientifically is a noteworthy topic.
In the work of [29], Gang Wang et al. proposed a FCANN framework, which
firstly clusters data using Fuzzy C-means algorithm, and then classifies each cluster
by ANN. This framework integrates unsupervised learning and supervised learning
to shorten the modelling time, alleviate the data imbalance problem and increase
52 3 Intelligent Network Awareness

the detection rate. In literature [30], Prasanta Gogoi et al. proposed a MLH-IDS
framework which has three layers, including a supervised layer, an unsupervised
layer and an outlier-based layer. The supervised layer is used to detect DoS [14]
and Probe [14] attacks, and the unsupervised layer is used to detect Normal data.
The outlier-based layer is used to distinguish R2L [14] and U2R [14] attacks from
each other. The hybrid multi-level framework takes full advantage of different ML
algorithms, which is more flexible and has a better performance.
An appropriate scheme of data engineering can also improve the performance of
IDS. Data engineering is an indispensable procedure in data mining which includes
some widely used techniques such as data preprocessing and dimension reduction
[32]. Data preprocessing techniques such as data cleaning and normalization can
help remove ‘dirty data’ and turn data into suitable form. The representative method
of dimension reduction is feature selection which is used to remove interfering and
redundant features to improve the performance of data mining project. In IDS, the
‘intrusion data’ is usually not suitable to be detected directly, which needs to be
processed by an appropriate data engineering method. Most works we mentioned
above mainly focus on the combination of ML algorithms without an elaborately
designed data engineering method.
In this part, we propose a novel IDS framework named HMLD through jointly
considering data engineering and machine learning. HMLD consists of three
modules, including Multi-Level Hybrid Data Engineering (MH-DE) module, Multi-
Level Hybrid Machine Learning (MH-ML) module, and Micro Expert Modify
(MEM) module. The MH-DE module focuses on data engineering and the MH-ML
module focuses on machine learning. These two modules construct a closed cycle
of Hybrid Multi-Level Data Mining, which provides a separated and customized
detection of different attacks. The hierarchical architecture can address the problems
caused by multicomponent and data imbalance. After performing the procedure of
MH-DE and MH-ML, most easily detected attacks have been marked. MEM module
is used to identify those new attacks which are difficult to detect. The HMLD
framework can be implemented in a variety of networks including mobile networks
by using different algorithms and parameters.
In this part, we use KDDCUP99 dataset to evaluate the performance of HMLD.
Experimental results show that HMLD can achieve 96.70% accuracy which is nearly
1% higher than the recent proposed optimal algorithm SVM+ELM+Modified K-
Means [13]. Meanwhile, it has a better performance than some other methods in
identifying DoS and R2L attacks.

3.2.1 The Framework of HMLD

In this subsection, we will introduce the framework and workflow of HMLD which
can detect different categories of attacks separately by different data engineering
methods and machine learning methods. The framework of HMLD is illustrated
in Fig. 3.11. The input of HMLD contains two parts, one is the labeled training
 Multi-Level Hybrid Machine Learning (MH-ML) module Micro Expert Modify
(MEM) module

Dtrain

Data
Proprecessed

Preprocessing
Attack1 _ DS Attack2 _ DS ... AttackN _ DS
Train Data
Modify
Training phase

Modify Model
M1 M2 ... MN Set

Abstract p%

Attack1 Attack2 AttackN Impurity Remaining

Filter Rest
Filter
... Filter Data Data
(1-p)%
Ddetect

Data
Remaining

Preprocessing
Attack1 Attack2 ... AttackN Predicted
3.2 Intrusion Detection Based on Hybrid Multi-Level Data Mining

Data
Detecing phase

Multi-Level Hybrid Data

Engineering (MH-DE) Ddetect_Label
module Detection and Evaluation Module

Fig. 3.11 Framework of HMLD

53
54 3 Intelligent Network Awareness

dataset Dtrain which is used to train the ML model, the other one is the unlabeled
detecting dataset Ddetect which is waiting to be detected. Dtrain is processed by
three modules: MH-DE, MH-ML and MEM as shown in Fig. 3.1, to construct
models that are used to detect intrusions in Ddetect. The output of HMLD is the
labeled detecting dataset Ddetect_Label. Algorithm 5 is the pseudo code of HMLD
which illustrates the step-by-step workflow of HMLD. We assume that there are N
attack categories in input data. The set of attacks in category i, where i ∈ [1, N ], is
denoted as Si . Each attack category has a corresponding package which is denoted
as Attacki _DS. We define the data engineering method used on Attacki _DS as
Di , and the ML model trained by Attacki _DS is denoted by Mi . P_key , P_nonkey
and Pi are intermediate variables in Algorithm 5.

Algorithm 5 HMLD
1: Input: Dtrain, Ddetect
2: Output: Ddetect_Label
3: Initialization: D = {Di , D2 , · · · , DN }, Ddetect_Label = ∅, P0 = Dtrain
4: MH-DE module
5: for i ∈ [1, N ] do
6: P_key = [data.label = i|data ∈ Si ]
7: P_nonkey = [data.label = 0|data ∈ / Si &data ∈ Pi−1 ]
8: Attacki _DS = do Di to (P_key + P_nonkey )
9: Pi = Pi − P_key
10: end for
11: MH-ML module
12: for i ∈ [1, N ] do
13: Mi = the ML model trained using Attacki _DS
14: Use Mi to detect Ddetect and label the detected attacks as i
15: Ddetect_Label = Ddetect_Label + detected attacks with label i
16: Ddetect = Ddetect − detected attacks with label i
17: end for
18: MEM module
19: Extract p% data from the remaining data in Ddetect and mark them by experts forming
modify set
20: Merge modify set and Dtrain to retrain a ML model to detect the rest of Ddetect return
Ddetect_Label

Dtrain is firstly sent to MH-DE module. The 4–10 lines in Algorithm 5 show the
workflow of MH-DE module. The line 5 means we will build packages and using
data engineering techniques to attacks in category i where i ∈ [1, N ] one by one.
The 6–8 lines in Algorithm 5 show how to build package Attacki _DS. We label
the attacks belonging to Si with i in Dtrain and denote this part of data as P_key .
Then we label remaining data which is belong to Pi−1 but not belong to Si with 0
and denote this part of data as P_nonkey . Thus package Attacki _DS is comprised
of P_key and P_nonkey by using Di to them which is shown in line 8. This step can
make the packages be more suitable for ML modelling by converting format and
removing redundancy. Package Pi is constructed by filtering out P_key . We repeat
this labeling process to build N packets in sequence.
3.2 Intrusion Detection Based on Hybrid Multi-Level Data Mining 55

After MH-DE, these packages are sent to MH-ML module. The 11–17 lines in
Algorithm 5 show the workflow of MH-ML module. In MH-ML, each package is
used as a training dataset to train an appropriate ML model as shown in line 13 in
Algorithm 5. The trained model Mi aims to correctly detect attacks in category i
as much as possible. In detecting phase, we use M1 , M2 , · · · , MN one by one to
mark and filter out attacks in different categories from Ddetect which is showed in
line 14–16 in Algorithm 5.
After this filtering procedure, the remaining unmarked data in Ddetect are
named impurity data. Impurity data mixes a large amount of normal data with
some difficult detected unknown attacks. The 18–20 lines in Algorithm 5 show the
workflow of MEM module. We send impurity data to MEM module. MEM module
samples micro data from the impurity data and marks them by experts to form a
modify set. Then MEM module merges the modify set with Dtrain to train a new
model to identify the unknown attacks in the impurity data. After finishing all the
procedures, we can get Ddetect_Label, which is the result of our detecting work.

3.2.2 HMLD with KDDCUP99

The HMLD framework can be applied in many different types of networks by

using different algorithms and parameters. In this part, we evaluate the performance
of the HMLD framework via KDDCUP99 [14] dataset. This section describes
the selection details of algorithms and parameters of each module when using
KDDCUP99.

3.2.2.1 KDDCUP99 Dataset

Firstly, we give a brief description of KDDCUP99 dataset. KDDCUP99 Intrusion

Detection Dataset is a classic dataset, which has 41 features and 5 types of labels
[52], namely Normal, DoS, Probe, U2R, and R2L, respectively. The meaning of the
labels are summarized in Table 3.7:

Table 3.7 The definitions of five labels

Label Meaning
Normal Normal data
DoS Denial-of-service
Probe Surveillance and probing
R2L Unauthorized access from a remote machine to a local machine
U2R Unauthorized access to local super-user privileges by a local machine
56 3 Intelligent Network Awareness

Table 3.8 The distribution of training data and testing data

Dataset Normal DoS Probe R2L U2R Sum
Training data 9725 39,164 417 111 3 49,420
Testing data 60,593 229,853 4166 16,189 228 311,028

In this part, we extract 10% data from kddcup.data_10_percent.gz [14] as

training data, and use corrected.gz [14] as testing data. Table 3.8 shows the number
of each category of attacks.
DoS, Probe, U2R and R2L are four attack categories. Each of them can be further
separated into many subcategories. There are 22 subcategories in the training data
and 39 subcategories in the testing data. The 17 new subcategories can be considered
as new or unknown attack subcategories. Among the four types of attacks, R2L
and U2R are much more difficult to be detected than DoS and Probe [30]. R2L is
difficult to be detected because it includes many new subcategories of attacks. U2R
is difficult to detect due to its sparseness.

3.2.2.2 MH-DE Module

MH-DE module focuses on data engineering which contains the stage of basic data
preprocessing and the stage of hybrid feature selection. Basic data preprocessing is
used to make the data suitable for modelling and hybrid feature selection is used to
remove redundant features.
We design the process of basic data preprocessing according to the feature of
KDDCUP99 dataset. There are three types of features in KDDCUP99, including
factorial type, continuous numerical type and discrete type. Therefore, the basic
data preprocessing for KDDCUP99 includes the factorial feature digitizing pro-
cedure and the continuous feature normalizing procedure. The former one will
map the factorial features into numbers. Without this digital procedure, the dataset
cannot be trained by a ML algorithm. The continuous feature normalizing procedure
normalizes all features to the range of [0, 1]. This step can eliminate the effect
caused by the diverse range of features.
In the stage of hybrid feature selection, we adopt different feature selection
methods for different packages according to the category of attacks. The flow chart
of MH-DE is demonstrated in Fig. 3.12. We number the attacks in category DoS,
Probe, U2R and R2L as 1, 2, 3, 4, respectively. In MH-DE module, we firstly pick
out all the DoS attacks and label them with 1. At the same time, we name the set
of other data as ‘Other1’ and label them as 0. Then do feature selection work to
the data. The selected features can distinguish DoS attacks from other data to the
utmost. The authors of [34] gave the results of the optimal feature selection subsets
of KDDCUP99 dataset, which is depicted in Table 3.9. The numbers in Table 3.9
represent the indexes of features in KDDCUP99. We use DoS feature selection
subset in Table 3.9 as the feature selection subset to get DoS_DS. After that, we pick
3.2 Intrusion Detection Based on Hybrid Multi-Level Data Mining 57

Fig. 3.12 Flow chart of MH-DE

Table 3.9 FMIFS feature selection results of KDDCUP99

Attack Feature selection subset generated by FMIFS
DoS 2,3,5,6,8,12,23,24,31,32,36,37
Probe 3,4,5,17,19,22,24,25,27,28,29,30,32,33,34,35,37,40,41
U2R 1,2,3,4,6,7,8,12,13,14,15,16,17,18,19,20,21,22,29,31,32,37,40
R2L 1,3,5,6,8,9,10,11,15,17,22,23,24,32,33
Overall 2,3,5,6,9,12,17,23,24,26,29,31,32,33,34,35,36,37,39

out all the Probe attack data from ‘Other1’ and label them with 2. In the meanwhile,
we name the remaining data as ‘Other2’ and label them with 0. Then we use Probe
feature selection subset in Table 3.9 to get Probe_DS. We repeat the this procedure
for U2R and R2L, respectively. After finishing the procedure of MH-DE module,
these five packages namely DoS_DS, Probe_DS, U2R_DS, R2L_DS and Train_DS
are formed for subsequent training.

3.2.2.3 MH-ML Module

After finishing the MH-DE module, the aforementioned five packages are sent into
MH-ML module. Each package needs to build a model to filter out its corresponding
category of attacks as many as possible. The authors of [29] proposed a modelling
framework including a clustering phase and a classifying phase. In HMLD, we adopt
this framework as our modelling framework because it can shorten the modelling
time and alleviate data imbalance problem. The basic model building process is
shown in Fig. 3.13. In this framework, the training data is firstly clustered by an
unsupervised clustering algorithm to get many clusters. Towards each cluster, we
train a specific supervised ML classifier.
The selection of algorithms and parameters of MH-ML module needs to be elab-
orate designed. We will use experiments to choose the algorithms and parameters
for MH-ML when using KDDCUP99 dataset.
58 3 Intelligent Network Awareness

Fig. 3.13 Framework of model building in MH-ML

Fig. 3.14 Different k for DoS_DS

In clustering phase, we adopt the K-Means [24] algorithm thanks to its good
performance and fast computation speed. The main idea of K-Means is clustering
data into several clusters according to their similarity. We define the number of
clusters as k, which has an impact on the performance of HMLD. Figures 3.14,
3.15, 3.16, and 3.17 show the Precision of HMLD with k ranging from 0 to 50 for
package DoS_DS, Probe_DS, U2R_DS, R2L_DS, respectively. The Precision of
detecting attacks represents the proportion of predicted attacks which are actually
attacks. When k is 0, it means that we do not use the K-Means. DoS attacks can reach
highest Precision when k is equal to 30 and 40. Therefore, we set k to 30 for DoS
attacks because a smaller k can reduce computing resources and shorten modelling
time. Probe attacks can achieve 91.91% Precision when k is 0 and 92.08% when k
is 20. Though the Precision is a bit lower when k is 0, the modelling complexity can
3.2 Intrusion Detection Based on Hybrid Multi-Level Data Mining 59

Fig. 3.15 Different k for Probe_DS

Fig. 3.16 Different k for U2R_DS

be significantly reduced. Hence we set k to 0 for Probe attacks. We set k to 20 and

30 for U2R and R2L attacks respectively to maximize the performance, which can
be observed from Figs. 3.16 and 3.17.
In classifying phase, we compare many ML algorithms including SVM[19],
ANN [20], DT[21] and RF[22] with different parameters. These algorithms are all
supervised learning algorithms which are mainly used in classification problems.
The main idea of linear SVM is to find a support hyperplane which can separate
different types of data in best performance. If the data can be separated much
better using a hypersurface than a hyperplane, we can use kernel SVM to replace
linear SVM. The most frequently-used kernel of SVM is rbf kernel, which has
another name as Gaussian kernel. Parameter r is the variance of the rbf kernel.
Parameter C is the relaxation factor of SVM [56]. Smaller C will cause a decline
of detection accuracy but can alleviate the overfitting problem. ANN is a multilayer
60 3 Intelligent Network Awareness

Fig. 3.17 Different k for R2L_DS

neural network which contains an input layer, many hidden layers and an output
layer. Each layer consists of many neurons which contains many parameters such
as weights, bias and activation functions. The activation functions can be identity,
logistic, tanh or relu. DT algorithm computes the information gain of each feature
and selects the biggest one as the root of a tree. Then repeat this procedure iteratively
until the stop condition is satisfied. CART is a representative algorithm of DT. RF
is a integration of DT, which samples data and features many times to build many
trees. Then RF gets the final results by comprehensively considering all these trees.
Figures 3.18, 3.19, 3.20, and 3.21 show the number of correctly detected attacks
by using different algorithms with different parameters. From these figures, we can
see that the performance of different algorithms varies greatly from each other for
different attacks. We choose the appropriate algorithm according to two metrics,
the number of detected attacks and the Precision of detecting attacks. The number
of detected attacks can be observed from Figs. 3.18, 3.19, 3.20, and 3.21. A better
algorithm can detect more attacks and can get a larger Precision.
For DoS attacks, SVM-linear (C = 1.0), ANN-identity, ANN-logistic and ANN-
tanh all can detect more attacks than other algorithms. The Precision of DoS
attack is 99.20% when using SVM-linear which is the highest among these four
algorithms. Hence, we select SVM-linear (C = 1.0) as the classification algorithm
of DoS_DS. For Probe attacks, ANN-logistic, CART can detect more attacks than
others. Comparing their Precision, CART is 69.40% and ANN -logistic is 92.62%.
For Probe attack, ANN-logistic algorithm is better. For U2R attacks, ANN-tanh
and ANN-identity can detect more attacks than others. When using ANN-tanh,
the Precision of U2R is 7.28%. But when using ANN-identity, the Precision is
increased to 37.14%. Thus, we choose ANN-identity for U2R_DS. For R2L attacks,
experiments show that SVM-rbf and ANN-relu have a better performance. However,
the Precision is 84.2% when using ANN-relu, and 82.26% when using SVM-rbf.
Therefore, we use ANN-relu algorithm for modelling of R2L_DS. The design of
MH-ML module is shown in Fig. 3.22.
3.2 Intrusion Detection Based on Hybrid Multi-Level Data Mining 61

Random Forest-100 trees 191923

Random Forest-50 trees 223092
Random Forest-20 trees 223112
Random Forest-10 trees 191919
Random Forest-5 trees 191917
CART 189465
ANN-relu 223108
ANN-tanh 223163
ANN-logistic 223161
ANN-identity 223179
SVM-rbf(C=10.0,r=100) 222259
SVM-rbf(C=10.0,r=50) 222343
SVM-rbf(C=10.0,r=5) 223079
SVM-rbf(C=10.0,r=1) 223078
SVM-rbf(C=10.0,r=0.1) 223073
SVM-rbf(C=5.0,r=100) 222256
SVM-rbf(C=5.0,r=50) 222338
SVM-rbf(C=5.0,r=5) 223072
SVM-rbf(C=5.0,r=1) 223078
SVM-rbf(C=5.0,r=0.1) 223118
SVM-rbf(C=1.0,r=100) 222209
SVM-rbf(C=1.0,r=50) 222273
SVM-rbf(C=1.0,r=5) 223074
SVM-rbf(C=1.0,r=1) 223071
SVM-rbf(C=1.0,r=0.1) 191792
SVM-linear(C=10.0) 223078
SVM-linear(C=5.0) 223073
SVM-linear(C=1.0) 223122
SVM-linear(C=0.1) 191792

180000 190000 200000 210000 220000 230000

Fig. 3.18 Correctly detected DoS attacks

3.2.2.4 MEM Module

After MH-ML, most of the attacks are marked and filtered out. We name the
remaining unmarked data impurity data. The impurity data mixes a large amount of
normal data with some unknown attacks. Before MEM, if we mark all the impurity
data as Normal, we can derive the ‘confusion matrix’ as Table 3.10. In a ‘confusion
62 3 Intelligent Network Awareness

Random Forest-100 trees 3161

Random Forest-50 trees 3050
Random Forest-20 trees 3069
Random Forest-10 trees 3062
Random Forest-5 trees 3299
CART 3707
ANN-relu
ANN-tanh 3047
ANN-logistic 3466
ANN-identity 2525
SVM-rbf(C=10.0,r=100) 2283
SVM-rbf(C=10.0,r=50) 2372
SVM-rbf(C=10.0,r=5) 2677
SVM-rbf(C=10.0,r=1) 3109
SVM-rbf(C=10.0,r=0.1) 3067
SVM-rbf(C=5.0,r=100) 2280
SVM-rbf(C=5.0,r=50) 2367
SVM-rbf(C=5.0,r=5) 2680
SVM-rbf(C=5.0,r=1) 3109
SVM-rbf(C=5.0,r=0.1) 2993
SVM-rbf(C=1.0,r=100) 2226
SVM-rbf(C=1.0,r=50) 2318
SVM-rbf(C=1.0,r=5) 2664
SVM-rbf(C=1.0,r=1) 3079
SVM-rbf(C=1.0,r=0.1) 2860
SVM-linear(C=10.0) 2610
SVM-linear(C=5.0) 2610
SVM-linear(C=1.0) 2603
SVM-linear(C=0.1) 2347
1500 2000 2500 3000 3500 4000

Fig. 3.19 Correctly detected probe attacks

matrix’, each row represents the number of data which is actually this type, and each
column represents data which is predicted as this type. For example, the number
in the upper left corner is the number of data which is actual normal and also be
predicted as normal. We can observe from Table 3.10 that a large amount of DoS
attacks and R2L attacks are wrongly detected before MEM module because there
3.2 Intrusion Detection Based on Hybrid Multi-Level Data Mining 63

Random Forest-100 trees 12

Random Forest-50 trees 7
Random Forest-20 trees 7
Random Forest-10 trees 2
Random Forest-5 trees 14
CART 17
ANN-relu 14
ANN-tanh 26
ANN-logistic 15
ANN-identity 26
SVM-rbf(C=10.0,r=50) 1
SVM-rbf(C=10.0,r=5) 8
SVM-rbf(C=10.0,r=1) 12
SVM-rbf(C=10.0,r=0.1) 11
SVM-rbf(C=5.0,r=50) 1
SVM-rbf(C=5.0,r=5) 8
SVM-rbf(C=5.0,r=1) 12
SVM-rbf(C=5.0,r=0.1) 11
SVM-rbf(C=1.0,r=5) 8
SVM-rbf(C=1.0,r=1) 11
SVM-linear(C=10.0) 11
SVM-linear(C=5.0) 11
SVM-linear(C=1.0) 11
0 5 10 15 20 25 30

Fig. 3.20 Correctly detected U2R attacks

appears many new subcategories of DoS and R2L attacks. Given that we have no
information about these new attacks in training data, these new attacks are difficult
to detect.
In order to detect these new attacks efficiently, we send impurity data to MEM
module. MEM module randomly samples p% of impurity data and marks them to
construct the modify set. We use DT [21] to retrain a ML model thanks to its rapid
modelling speed. This model is used to detect the new attacks in impurity data. We
experiment different value of p to compare the average accuracy of HMLD, and
the results are shown in Table 3.11. When p% is 0.3%, HMLD achieves a relative
high performance. When p% is larger than 0.3%, the growth starts to slow down.
Therefore, we set p to 0.3.
MEM module samples 0.3% data from impurity data and marks them by experts
to form the Modify Set. Experimental results show that there are about 240 records
in the Modify Set. After MEM, the ‘confusion matrix’ is showed in Table 3.12. With
this micro cost, most of the unknown attacks of DoS and R2L are correctly detected.
64 3 Intelligent Network Awareness

Fig. 3.21 Correctly detected R2L attacks

Fig. 3.22 Parameters and algorithms selection for MH-ML

3.2 Intrusion Detection Based on Hybrid Multi-Level Data Mining 65

Table 3.10 Confusion matrix before MEM

Actual vs predict Normal DoS Probe R2L U2R
Normal 59,391 711 202 272 17
DoS 6674 2,231,222 57 0 0
Probe 358 341 3466 0 1
R2L 13,911 617 28 1626 7
U2R 48 122 18 14 26

Table 3.11 Comparisons of different p

p 0.01% 0.05% 0.1% 0.2% 0.3% 0.4% 0.5%
Accuracy 92.56% 95.12% 96.00% 96.21% 96.50% 96.60% 96.62%

Table 3.12 Confusion matrix after MEM

Actual vs predict Normal DoS Probe R2L U2R
Normal 56,210 763 697 2699 39
DoS 192 229,565 70 6 0
Probe 34 481 3615 35 1
R2L 4333 620 86 11,102 9
U2R 28 122 32 20 26

3.2.3 Experimental Results and Discussions

3.2.3.1 Evaluation Criteria

TP(true positives), TN(true negatives), FP(false positives), FN(false negatives) [34]

are usually used to evaluate the performance of IDS. In IDS, we define normal data
as positive and define attacks as negative. TP is the number of data which is actual
positive and also predicted as positive. TN is the number of data which is actual
positive but predicted as negative. FP is the number of data which is actual negative
but predicted as positive. FN is the number of data which is actual negative and also
predicted as negative. TP, TN, FP and FN can be written in a ‘confusion matrix’ as
shown in Table 3.13.
Precision, Recall, F-value and Accuracy [34] are also defined to evaluate the
performance. Precision given by Eq. (3.11) is the proportion of predicted positives
values which are actually positive. Recall given by Eq. (3.12) is the proportion of
the actual number of positives which are correctly identified. Recall also can be
called as Detection rate. F-value given by Eq. (3.13) is a harmonic mean between
66 3 Intelligent Network Awareness

Table 3.13 Confusion matrix

Predict as positive Predict as negative
Actual is positive TP TN
Actual is negative FP FN

Precision and Recall. Accuracy given by Eq. (3.14) is the proportion of correctly
predicted data.

TP
P = P recision = , (3.11)
T P + FP

TP
R = Recall = Detection rate = , (3.12)
T P + FN

2 × P recision × Recall
F − value = , (3.13)
P recision + Recall

TP +TN
Accurary = . (3.14)
T P + T N + FN + FP

3.2.3.2 Experiments and Analysis

(1) Performance of HMLD

In this part, we show the overall performance of HMLD in terms of evaluation
indicators including Precision, Recall, F-value and Accuracy. Table 3.14 shows the
Precision, Recall and F-valve of each attack. The Accuracy is 96.70% which can be
computed by Eq. (3.14).
(2) Comparisons of Hybrid Feature Selection Used by ML-DE with Other
Feature Selection Methods
In the ML-DE module of HMLD, we adopt hybrid feature selection as the feature
selection method in data engineering. In contrast, we tested another two commonly
used feature selection methods. One is doing the same feature selection to all
packages, which called unified feature selection. When do unified feature selection,
we use the feature subset of Overall in Table 3.9 to all packages. The other is not
to do feature selection, which called no feature selection. Table 3.15 and Fig. 3.23
shows the comparisons of TP and Precision of different attacks by using hybrid
feature selection, unified feature selection and no feature selection.
We can see that the overall performance of hybrid feature selection is much better
than the other methods. From Table 3.15, the numbers of correctly detected attacks
using hybrid feature selection are the highest for DoS, Probe and R2L. For U2R, the
number is just 1 less than no feature selection. In terms of Precision, from Fig. 3.23
3.2 Intrusion Detection Based on Hybrid Multi-Level Data Mining 67

Table 3.14 Indicators of HMLD

Indicators(%) Normal DoS Probe R2L U2R Overall
Precision 92.46 99.14 80.33 80.10 34.67 96.55
Recall 93.05 99.88 86.77 68.74 11.40 96.70
F-value 92.75 99.51 83.42 73.98 17.16 96.60

Table 3.15 Numbers of correctly detected attacks

No. of correctly detected attacks DoS Probe U2R R2L
Hybrid feature selection 223,122 3466 26 1626
Unified feature selection 222,630 2662 4 748
No feature selection 208,424 3355 27 252

Fig. 3.23 Comparison of Precision for different attacks by using different feature selection
methods

we can see that, the Precision using hybrid feature selection is much higher than
the other two methods for Probe, R2L and U2R. For DoS, the Precision is almost in
the same level. The reason is that the characteristic of different attacks is reflected
in different subset of features. Using customized subset of features can improve the
performance on the corresponding attacks. Therefore, we can conclude that hybrid
feature selection is far superior to other methods.
(3) Comparisons of Hybrid Multi-Level ML Used by MH-ML with Single ML
Methods
In MH-ML of HMLD, we use the method of selecting different ML algorithms
for different packages according to their corresponding category of attacks. This
method is called Hybrid Multi-Level Machine Learning method. In the meanwhile,
we call the method of using the same ML algorithm such as SVM, ANN and RF
to all packages as Single Machine Learning method. Figure 3.24 contrasts the
performance of using Hybrid Multi-Level Machine Learning and Single Machine
Learning. From Fig. 3.24 we can observe that the Recall and F-value of using
Hybrid Multi-Level Machine Learning method is much better than using Single
68 3 Intelligent Network Awareness

Fig. 3.24 Performance comparison of Hybrid multi-level machine learning and Single machine
learning

Machine Learning methods. The Precision of using Hybrid Multi-Level Machine

Learning method is higher than using ANN and RF, but a bit less than using SVM.
Different algorithms has a different performance on detecting different attacks and
Hybrid Multi-Level Machine Learning choose suitable algorithms to each category
of attacks. Hence, the overall performance of using Hybrid Multi-Level Machine
Learning is much better than using Single Machine Learning.

(4) Comparisons of HMLD with Prior Works

In this subsection, we compare the performance of HMLD with some prior works
[13, 35, 37] proposed in recent years. The authors of literature [13] proposed a
modified K-Means to build a high-quality training dataset and train a multi-level
hybrid intrusion detection model using SVM and ELM. The authors of literature
[35] proposed an IDS based on SVM, hierarchical clustering algorithm and a simple
feature selection procedure. The literature [37] is the method proposed by the winner
of KDDCUP99 competition.
Tables 3.16 and 3.17 show comparisons between HMLD and these aforemen-
tioned works. In Table 3.16, each row represents the detection rates of different data
types generated by one algorithm. Each column represents the detection rates of one
data type by using different algorithms.
According to Table 3.16, HMLD has a high Detection rate of DoS and R2L
attacks. Though the Detection rate of Normal and Probe is not the highest, they have
the rates at or just below the average. The Detection rate of U2R is low, which is
because the number of U2R attacks in training data is too small. From Table 3.12, we
can see that most of the U2R attacks are predicted as DoS attacks. Though detected
as wrong category, it is still classified as an attack category, not a normal class. From
Table 3.17, we can observe that the overall performance of HMLD is better than the
prior works, with a detection Accuracy reaching 96.70%, which is nearly 1% higher
than the optimal in the other. In summary, the overall performance of HMLD model
is better than prior works.
3.3 Abnormal Network Traffic Detection Based on Big Data Analysis 69

Table 3.16 Detection rate (%) generated by HMLD and some prior works
Algorithms Normal DoS Probe R2L U2R
HMLD-KDD 93.05 99.88 86.77 68.74 11.40
SVM+ELM+Modify K-Means 98.13 99.54 87.22 21.93 31.79
SVM+BIRCH clustering 99.30 99.50 97.50 19.70 28.80
Winner of the KDDCUP99 99.50 97.10 83.30 13.20 8.40

Table 3.17 Accuracy (%) Algorithms Accuracy(%)

comparisons of HMLD and
some prior works HMLD-KDD 96.70
SVM+ELM+Modify K-means 95.75
SVM+BIRCH clustering 95.70
Winner of the KDDCUP99 93.30

3.3 Abnormal Network Traffic Detection Based on Big Data

Analysis

With the improvement of network, storage, calculation and transmission,the Internet

is interacting more closely with people than ever before. While the Internet is
making our life more convenient, it brings about some potential risks [6]. For
example, malicious attacks involving user privacy and security become more and
more frequent.
The changes of how people use the Internet is a new challenge for traditional
abnormal network event detection techniques. It is more hard for researchers to get
aware of some new kinds of attacks [62]. To resolve these problems, some abnormal
network traffic detection methods have been proposed. Traditional abnormal traffic
detection method can be classified into two categories [38, 41, 42, 47]. One is
misuse detection, and the other is abnormal detection. The two methods have their
own pros and cons. Misuse detection has a high accuracy but needs support from
known knowledge. Abnormal detection do not need known knowledge, but cannot
categorize the type of attacks, the accuracy is also lower. For example, Hari Om [44]
designes a hybrid detection system, which is a hybrid anomaly detection system
considering k-means, k-nearest neighbor and Naïve Bayes methods.
However, the explosive increase of network traffic has directly or indirectly
pushed the Internet into the big data era, which makes anomaly traffic detection
more difficult to deal with because of high calculation volume and constant changes
of network data distribution caused by big data [45, 46, 48, 50]. Because the speed of
network data generation is fast, it makes the volume of normal traffic and abnormal
traffic differ a lot, and the distribution of the data change. Besides, with big data,
the difference between normal traffic and abnormal traffic is increasing [55, 58]. It
makes the traditional methods unable to effectively detect abnormal traffic.
Therefore, to increase the accuracy of abnormal traffic and avoid the loose caused
by false negative detection, we propose a novel model based on big data analytics,
which can avoid the influence brought by adjustment of network traffic distribution,
70 3 Intelligent Network Awareness

increase dectection accuracy and reduce the false negative rate. The core of the
proposed model is not simply combination of traditional detection methods, but
a novel detection model based on big data. In the simulation, we use k-means,
decision tree and random forest algorithms as comparative objects to vertify the
effectiveness of our model. Simulation results reveal that the proposed model has a
much better performance, which can achieve a detection rate of 95.4% on normal
data, 98.6% on DoS attack, 93.9% on Probe attack, 56.1% on U2R attack, and 77.2%
on R2L attack.

3.3.1 System Model

Influenced by big data, network data distribution is gradually changing [5]. This part
try to solve the problem that caused by the increasing difference between normal
traffic and abnormal traffic. Therefore, we proposed a new abnormal traffic detection
model based on big data analysis, and this model includes three sub-models.

3.3.1.1 Normal Traffic Selection Model

Normal traffic selection model uses classification and clustering algorithm to

distinguish normal and anomaly behaviors, rather than involved specific anomaly
behaviors. This model includes two stages:
1. Training stage: training model uses data that labeled normal or abnormal, and the
model applies in test stage.
2. Test stage: test stage is similar to detection in practice. Using unlabeled date, the
model classifies traffic data into normal or abnormal, and labels them.
Normal traffic selection model uses k-means clustering algorithm, KNN, deci-
sion tree and random forest classification algorithms. Traditionally, before using
k-means algorithms, it is very important to set the number of categories, because we
don’t know how many categories. But in order to distinguish normal and abnormal
behavior, the normal traffic selection model uses k-means as following way.
In training stage, using labeled information classify data into normal and
abnormal. These two categories use k-means separately instead of clustering all
data at once, getting the center of the data set respectively. Then using the center of
the data set, KNN clustering algorithm classifies test data. Decision tree and random
forest classification algorithms train with labeled normal and abnormal data.
3.3 Abnormal Network Traffic Detection Based on Big Data Analysis 71

3.3.1.2 Abnormal Traffic Selection Model

The purpose of abnormal traffic selection model is avoid influence caused by too
many normal traffic than abnormal traffic. This model classifies anomaly traffic into
specific categories, and includes two stage as well:
1. Training stage: this stage only use abnormal data to train classification model,
and every data label specific attack group. Using classification algorithms learns
classified rules.
2. Test stage: test stage is similar to detection in practice, using unlabeled data
(including normal behavior data). The classification model classifies anomaly
traffic into specific categories according to the classified rules, and gives specific
label to every data.
Abnormal traffic selection model uses decision tree and random forest classifica-
tion algorithms. Abnormal traffic selection model and normal traffic selection model
are independent, without order of priority in training stage or test stage.

3.3.1.3 Abnormal Traffic Selection Model

Mixed compensation model combines the result from normal traffic selection model
and abnormal traffic selection model to produce a final result. Although abnormal
traffic selection model is more effective because without influence of normal traffic
data, the model has high false negative rate due to this characteristic. Therefore use
normal set produced by normal traffic selection model to compensate abnormal set
A = {A1 , A2 , · · · , Ak } produced by abnormal traffic selection model. Ai , i ∈ [1, k]
denote specific attack category. If c denote detection result, rule of compensation as
follow:

if c ∈ Ai , c ∈ N, then c ∈ N
(3.15)
if c ∈ Ai , c ∈
/ N, then c ∈ Ai .

3.3.2 Simulation Results and Discussions

Before using three sub-models of anomaly detection based on big data analysis,
data set needs be preprocessed with label for training model. It should be noted that
rightly selecting feature is a good way to reduce dimension and increase efficiency
of running. In the simulation, three different algorithms are used to verify validity
of the proposed model.
72 3 Intelligent Network Awareness

3.3.2.1 Data Set

In the simulation, we use KDDCUP99 [53] data set to test my model. KDDCUP99
data set is widespread use for testing abnormal detection model, which is obtained
and processed from KDDCUP99 [54]. KDDCUP99 data set has 41 features and
been sorted into three group: basic feature, content feature and time feature [57].
The distribution of data set is shown as Table 3.18, where training data has 5
million records, 10% of training data has 494,021 records, and test data has 311,029
records. Every record is labeled to be normal or abnormal, and abnormal data can
be classified into four groups: Dos, U2R, R2L and Probe. From Table 3.18, we find
that normal data in training data set is more than abnormal data in test data set.
Therefore, this data set can be used to test the performance of the proposed model
under different circumstances.

3.3.2.2 Simulation Results

As shown in Table 3.19, we have done eight experiments with the model based on
big data analysis, and three control experiments which used k-means, decision tree
or random forest respectively. In the control groups, training classify model uses all
training data set with five categories, then classifying test data into five categories.
Another control group is winner of KDDCUP99.

Table 3.18 Distribution of KDDCUP99 data set

Data set Normal DoS Probe R2L U2R
10% of training data set 97,278 391,458 4107 1126 52
Test data set 60,593 229,853 4166 16,189 228

Table 3.19 Number of experiments

Normal traffic Abnormal traffic
No. selection model selection model No. of control group Algorithm
1 k-means1a Random forest 9 k-means
2 k-means1a Decision tree 10 Decision tree
3 k-means2a Random forest 11 Random forest
4 k-means2a Decision tree 12 Winner of
KDDCUP99
5 Decision tree Decision tree
6 Decision tree Random forest
7 Random forest Decision tree
8 Random forest Random forest
a Inthe normal traffic selection model, the number of cluster of normal and abnormal respectively
is 4 and 30 in k-means1, and the number of cluster of normal and abnormal respectively is 100
and 300 in k-means2
3.3 Abnormal Network Traffic Detection Based on Big Data Analysis 73

Table 3.20 Prediction accuracy

No. Experiment Normal DoS Probe U2R R2L
1 k-means1+Random forest 0.632 0.814 0.939 0.561 0.679
2 k-means1+Decision tree 0.656 0.791 0.878 0.500 0.772
3 k-means2+Random forest 0.945 0.983 0.910 0.513 0.510
4 k-means2+Decision tree 0.946 0.979 0.852 0.500 0.504
5 Decision tree+Decision tree 0.951 0.984 0.829 0.500 0.512
6 Decision tree + Random forest 0.951 0.986 0.831 0.550 0.517
7 Random forest + Decision tree 0.954 0.980 0.861 0.500 0.521
8 Random forest + Random forest 0.952 0.985 0.872 0.520 0.510
9 k-means 0.938 0.968 0.785 0.500 0.528
10 Decision tree 0.951 0.983 0.793 0.500 0.500
11 Random forest 0.952 0.985 0.875 0.522 0.507
12 Winner of KDDCUP99 0.995 0.971 0.833 0.132 0.084

Table 3.21 Compared with winner of KDDCUP99

No. Experiment Normal DoS Probe U2R R2L Final score Rank
8 Random forest+Random forest 3 2 2 2 4 13 1
6 Decision tree+Random forest 4 1 6 1 2 14 2
7 Random forest +Decision tree 2 5 3 4 1 15 3
2 k-means2+Random forest 7 4 1 3 5 20 4
5 Decision tree+Decision tree 4 3 7 6 3 23 5
4 k-means2+Decision tree 6 6 4 5 6 27 6
12 Winner of KDDCUP99 1 7 5 7 7 27 6

In the simulation, prediction accuracy is used as a simulation metric of detection

effect, which is shown in Table 3.20. Besides, we adopt way of sorting and grading
for every type. For example, all experiments are sorted by prediction accuracy of
normal. The first grades 1 point, the second grades 2 points, and so on. Finally,
adding grade of five groups is final grade.
As shown in Table 3.21, the experiment group and winner of KDDCUP99 are
sorted by final grade. While the later has high detection rate in normal data, as for
four attack types, the result of model based on big data analysis is better than winner
of KDDCUP99.
Algorithm of winner of KDDCUP99 is C5 decision tree [59–61, 63]. Training
data of winner of KDDCUP99 is a little different with my experiment. Thus for
evaluating detection effect of the proposed mode, we did three control experiments
with same training data and test data, used with k-means, decision tree or random
forest respectively. The number of these experiments is noted as 11, 10 and 9.
Sorting result shows that detection effect of algorithm of the proposed model is
better than no use, as shown as Table 3.22. We will discuss experiments results,
compared No. 8 with No. 11, No. 7 with No. 5 and No. 3 with No. 4.
74 3 Intelligent Network Awareness

Table 3.22 Compared with control group

No. Experiment Normal DoS Probe U2R R2L Final score Rank
6 Decision tree+Random forest 4 1 6 1 3 15 1
8 Random forest + Random forest 2 2 3 3 5 15 1
11 Random forest 2 2 2 2 7 15 1
7 Random forest + Decision tree 1 7 4 5 2 19 4
3 k-means2+ Random forest 8 5 1 4 5 23 5
5 Decision tree + Decision tree 4 4 7 5 4 24 6
10 Decision tree 4 5 8 5 9 31 7
9 k-means 9 9 9 5 1 33 8
4 k-means2+ Decision tree 7 8 5 5 8 33 8

3.3.2.3 Discussing Result of No. 8 and No. 11

Score of top three are same. Judging No. 8 and No. 11 with final grade, detection
result of two experiments are almost same. And both of them use random forest
algorithm. But the difference is:
1. Importance of variable used in classifying is different;
2. No. 8 has lower false negative rate.
• Importance of Variable
As shown as Fig. 3.25, variables chosen by random forest in No. 8 and No. 11 are
different. Random forest algorithm can output importance of variables, noted Gini
index [51]. Figure 3.25 shows that top 20 have important variables in comparison
with top 1, whose value is higher and more important.
In No. 8, rank of variables is different between normal traffic selection model
and abnormal traffic selection model. This means that variable used for predicting
normal or abnormal and specific attack is different. Therefore, choosing variable in
No. 11 is influenced by both sides, and output a compromised result when choosing
variables, that’s why prediction of model in No. 11 has deviation.
• Comparison of False Negative Rate
In order to evaluate effect on predicting abnormal behavior, false negative rate is
used as an important index, which can measure how many attack events are omitted.
Table 3.23 shows confusion matrix of results of experiments No. 8 and No. 11 when
using random forest. Row express information of prediction, and column express
actual information. False negative rate of No. 8 in normal type is very low, but
high in U2R and R2L type. In No. 8, false negative rate of normal selection model
in normal is low. Without influence of normal training data, false negative rate of
abnormal selection model in four specific attack types are lower than No. 11.
3.3 Abnormal Network Traffic Detection Based on Big Data Analysis 75

Fig. 3.25 Importance of variables in random forest. (a) No. 11. (b) No. 8 normal traffic selection
model. (c) No. 8 abnormal traffic selection model
76 3 Intelligent Network Awareness

Fig. 3.25 (continued)

3.3.2.4 Discussing Result of No. 5 and No. 7

No. 5 and No. 7 respectively compare with No. 6 and No. 8 by using same algorithm
in normal traffic selection model, and their ranks are lower when using decision tree
in abnormal traffic selection model.
Table 3.24 is confusion matrix of abnormal traffic selection model with decision
tree algorithm. It shows that U2R can not be detected and false negative rate of R2L
is higher. In order to find the reason, classify tree is checked in Fig. 3.26, where the
classification model prefers DoS and Probe attack, then R2L attack, and no result
point of U2R attack. Distribution of training data can explain this phenomenon,
which can be shown in Fig. 3.27.
When generating decision tree, the obtained information will cause results in
favor of feature which have more samples. Therefore, if the number of training
data set in every group is different enough, it cannot get efficient classification
model for small samples. Moreover, because the number of between training data
is comparatively equal, classification result is better, such as No. 6, when normal
traffic selection model uses decision tree.
3.3 Abnormal Network Traffic Detection Based on Big Data Analysis 77

Table 3.23 Confusion matrix

No. 11
Prediction Normal DoS Probe U2R R2L
Normal 60,287 5967 847 159 15,839
DoS 69 223,814 191 8 0
Probe 233 72 3128 50 104
U2R 1 0 0 10 5
R2L 3 0 0 1 241
False negative 0.00505 0.026273 0.24916 0.95614 0.985113
No. 8 normal traffic selection model
Prediction Normal Abnormal
Normal 60,289 22,853
Abnormal 304 227,583
False negative 0.005017 0.091253
No. 8 abnormal traffic selection model
Prediction DoS Probe U2R R2L
DoS 229,231 769 20 4693
Probe 297 3393 135 5646
U2R 0 0 39 32
R2L 325 4 34 5818
False negative 0.002706 0.18555 0.828947 0.64062

Table 3.24 Confusion matrix of abnormal traffic selection model with decision tree
Predition DoS Probe U2R R2L
DoS 227,792 589 34 6245
Probe 1434 3192 20 283
U2R 0 0 0 0
R2L 627 385 174 9661

3.3.2.5 Discussing Result of No. 3 and No. 4

No. 3 and No. 4 use k-means in normal traffic selection model to choose clustering
center. Table 3.25 shows final prediction accuracies in No. 3 and No. 4. Because
final results are lower than that of normal traffic selection model or abnormal
traffic selection model, we find that this problem is caused by using k-means
in normal selection model. Table 3.26 shows confusion matrix of normal traffic
selection model of No. 3 and No. 4. Many abnormal records are predicted as
normal, which cause high false negative rate. Therefore, many abnormal records
predicted by abnormal traffic selection model will be regarded as normal after mixed
compensation model.
78 3 Intelligent Network Awareness

yes dst_host < 0.12 no

dos
5285 / 396743

count >= 1.5

probe
dos 431 / 3175
2493 / 393568

service < 12

dos probe
435 / 391197 1101 / 2371

probe r21
0 / 1250 369 / 1121

Fig. 3.26 Classify tree of abnormal traffic selection model

Fig. 3.27 Distribution of training data

Table 3.25 Accuracy of No. 3 and No. 4

No. Model Algorithm Accuracy
No. 3 Normal traffic selection k-means 0.926
Abnormal traffic selection Random forest 0.948
Mixed compensation model 0.923
No. 4 Normal traffic selection k-means 0.925
Abnormal traffic selection Decision tree 0.961
Mixed compensation model 0.918
3.3 Abnormal Network Traffic Detection Based on Big Data Analysis 79

Table 3.26 Confusion matrix of normal traffic selection model of No. 3 and No. 4
No. Prediction Normal Abnormal
No. 3 Normal 59,189 21,663
Abnormal 1404 228,773
No. 4 Normal 59,428 22,221
Abnormal 1165 228,215

Table 3.27 Results of experiments

No. Experiment DoS Probe U2R R2L Final Rank
1 k-means1+Random forest 10 1 1 2 14 1
6 Decision tree+ Random forest 1 8 2 5 16 2
8 Random forest + Random forest 2 5 4 7 18 3
11 Random forest 2 4 3 9 18 3
3 k-means2+ Random forest 5 2 5 7 19 5
2 k-means1+ Decision tree 11 3 6 1 21 6
7 Random forest + Decision tree 7 6 6 4 23 7
5 Decision tree + Decision tree 4 9 6 6 25 8
9 k-means 9 11 6 3 29 9
4 k-means2+ Decision tree 8 7 6 10 31 10
10 Decision tree 5 10 6 11 32 11

Nowadays, many novel attacks are unknown to researchers, and many attacks
will be disguised as normal. It’s very dangerous to have high false negative rate, and
it does not fit the proposed model.
Because the effect of k-means has great correlation with the number of centers
chosen to cluster, and we can fine tune the strength of clustering, and lower the false
negative rate to establish a strict normal selection model.
In No. 3 and No. 4, the number of centers for normal traffic and attacks is 100 and
300, respectively. Although it can achieve a good overall accuracy, its false negative
rate is higher than other model. However, according to Table 3.27, by choosing 4
and 30 in No. 1 and No. 2, it has lower false negative rate, and only classify four
kinds of attacks. Besides, a strict normal detection model is established.
By adjusting the parameters and reducing false negative rate in No. 1 and No.
2, we can find that the rank has increased rapidly compared with No. 3 and No. 4.
Especially, when K-means combines with random forest, it has a very high accuracy
on Probe, U2R and R2L attack. Therefore, we can draw the conclusion that by
adjusting the parameters of K-means, the strength of abnormal traffic detection can
be controlled by adjusting the strength of normal traffic identification.
80 3 Intelligent Network Awareness

Table 3.28 Summary of model

Model 1 Model 2 Model 3
Normal traffic selection model k-means1 Decision tree Random forest
Abnormal traffic selection model Random forest Random forest Random forest

Based on the results analyzed above, as shown in Table 3.28, the following
conclusions can be drawn:
1. Random forest classification algorithm can adapt to the change of distribution of
network data, and this algorithm by using the proposed model can reduce false
negative rate.
2. If the number of training data in different group is largely different with each
other, the classify model built by decision tree will prefer to attack types, which
have more training data. So we should avoid using decision tree in abnormal
traffic selection model. However, in the normal traffic selection model, the
difference between different groups is comparatively small. In this situation,
using decision tree can fast get classify model, and the results have higher
accuracy.
3. There are more and more unknown abnormal events in the future. In order to
avoid loss of false negative prediction, we can change the number of clustering
in the normal traffic selection model with k-means algorithm to reduce false
negative rate and increase the accuracy of detecting abnormal events.

3.4 Summary

In this chapter, we discuss the main challenge of network traffic intelligent aware-
ness and introduced several machine learning based traffic awareness algorithms.
we first present a multi-level intrusion detection model framework named MSML
to address these issues. Then, we propose a novel IDS framework called HMLD
to address these issues, which is an exquisite designed framework based on Hybrid
Multi-Level Data Mining. In addition, we propose a new model based on big data
analysis, which can avoid the influence brought by adjustment of network traffic
distribution, increase detection accuracy and reduce the false negative rate. Finally,
we proposes an end-to-end IoT traffic classification method relying on a deep
learning aided capsule network for the sake of forming an efficient classification
mechanism that integrates feature extraction, feature selection and classification
model.
References 81

References

1. J. Du, C. Jiang, H. Zhang, X. Wang, and M. Debbah, “Secure satellite-terrestrial transmission

over incumbent terrestrial networks via cooperative beamforming,” IEEE Journal on Selected
Areas in Communications, vol. PP, no. 99, pp. 1–1, 2018.
2. C.-H. Tsang, S. Kwong, and H. Wang, “Genetic-fuzzy rule mining approach and evaluation
of feature selection techniques for anomaly intrusion detection,” Pattern Recognition, vol. 40,
no. 9, pp. 2373–2391, 2007. [Online]. Available: http://www.sciencedirect.com/science/
article/pii/S0031320306005218
3. J. Wang, C. Jiang, H. Zhu, R. Yong, and L. Hanzo, “Internet of vehicles: Sensing aided trans-
portation information collection and diffusion,” IEEE Transactions on Vehicular Technology,
vol. 67, no. 5, pp. 3813–3825, 2018.
4. J. Wang, C. Jiang, H. Zhang, Z. Xin, V. C. M. Leung, and L. Hanzo, “Learning-aided network
association for hybrid indoor lifi-wifi systems,” IEEE Transactions on Vehicular Technology,
vol. PP, no. 99, pp. 1–1, 2017.
5. J. Du, C. Jiang, K. C. Chen, R. Yong, H. V. Poor, J. Du, C. Jiang, K. C. Chen, R. Yong,
and H. V. Poor, “Community-structured evolutionary game for privacy protection in social
networks,” IEEE Transactions on Information Forensics & Security, vol. 13, no. 3, pp. 574–
589, 2018.
6. X. Zhu, C. Jiang, L. Yin, L. Kuang, G. Ning, and J. Lu, “Cooperative multigroup multicast
transmission in integrated terrestrial-satellite networks,” IEEE Journal on Selected Areas in
Communications, vol. PP, no. 99, pp. 1–1, 2018.
7. J. Erman, A. Mahanti, M. Arlitt, I. Cohen, and C. Williamson, Offline/realtime traffic
classification using semi-supervised learning, Performance Evaluation, vol. 64, no. 9–12, pp.
1194–1213, 2007.
8. H. Yao, D. Fu, P. Zhang, M. Li, and Y. Liu, “Msml: A novel multi-level semi-supervised
machine learning framework for intrusion detection system,” IEEE Internet of Things Journal,
2018.
9. J. Zhang, X. Chen, Y. Xiang, W. Zhou, and J. Wu, “Robust network traffic classification,”
IEEE/ACM Transactions on Networking, vol. 23, no. 4, pp. 1257–1270, 2015.
10. J. Davis and M. Goadrich, “The relationship between precision-recall and roc curves,” in
International Conference on Machine Learning, 2006, pp. 233–240.
11. L. He, “An improved intrusion detection based on neural network and fuzzy algorithm,”
Journal of Networks,9,5(2014-05-08), vol. 9, no. 5, 2014.
12. W. Xuren, H. Famei, and X. Rongsheng, “Modeling intrusion detection system by discovering
association rule in rough set theory framework,” in International Conference on Computa-
tional Intelligence for Modelling, Control & Automation, 2006, p. 24.
13. W. L. Al-Yaseen, Z. A. Othman, M. Z. A. Nazri, Multi-level hybrid support vector machine
and extreme learning machine based on modified K-means for intrusion detection system,
Expert Systems with Applications 67 (2017) 296–303.
14. KDD Cup 1999 Data, http://kdd.ics.uci.edu/databases/kddcup99/kddcup99.html.
15. Q. M. Alriyami, E. Asimakopoulou, N. Bessis, A Survey of Intrusion Detection Systems
for Mobile Ad Hoc Networks, in: International Conference on Intelligent Networking and
Collaborative Systems(INCoS), Salerno, Italy, 2014.
16. H. Yao, L. Chong, P. Zhang, and L. Wang, “A feature selection method based on synonym
merging in text classification system,” Eurasip Journal on Wireless Communications &
Networking, vol. 2017, no. 1, p. 166, 2017.
17. D. E. Denning, An Intrusion-Detection Model, in: IEEE Symposium on Security and Privacy,
Oakland, CA, USA, 1986.
18. Y. He, F. R. Yu, N. Zhao, H. Yin, H. Yao, R. C. Qiu, Big Data Analytics in Mobile Cellular
Networks, IEEE Access 4 (2016) 1985–1996.
19. V. Vapnik, C. Cortes, Support vector networks, Machine Learning 20 (3) (1995) 273–297.
82 3 Intelligent Network Awareness

20. G. Zhang, B. E. Patuwo, M. Y. Hu, Forecasting with artificial neural networks:: The state of
the art, International Journal of Forecasting 14 (1) (1998) 35–62.
21. J. R. Quinlan, Induction of Decision Trees, Machine Learning 1 (1) (1986) 81–106.
22. A. Cutler, D. R. Cutler, J. R. Stevens, Random Forests, Machine Learning 45 (1) (2004) 157–
176.
23. A. K. Jain, M. N. Murty, P. J. Flynn, Data clustering: a review, ACM Computing Sur-
veys(CSUR) 31 (3) (1999) 264–323.
24. J. A. Hartigan, M. A. Wong, A K-means Clustering Algorithm, Applied Statistics 28 (1)
(1979) 100–108.
25. P. Zhang, S. Wu, M. Wang, H. Yao, and Y. Liu, “Topology based reliable virtual network
embedding from a qoe perspective,” China Communications, vol. 15, no. 10, pp. 38–50, 2018.
26. K. Khan, S. U. Rehman, K. Aziz, S. Fong, S. Sarasvady, DBSCAN: Past, present and
future, in: Fifth International Conference on the Applications of Digital Information and Web
Technologies (ICADIWT), Bangalore, India, 2014.
27. K. Wang, J. Zhang, D. Li, X. Zhang, T. Guo, Adaptive Affinity Propagation Clustering, Acta
Automatica Sinica 33 (12) (2007) 1242–1246.
28. W. Wu, H. Yao, T. Huang, L. Wang, Y. Zhang, and Y. Liu, “Survey of development on future
networks and industrial internet,” Journal of Beijing University of Technology, vol. 43, no. 2,
pp. 163–172, 2017.
29. G. Wang, J. Hao, J. Ma, L. Huang, A new approach to intrusion detection using Artificial
Neural Networks and fuzzy clustering, Expert Systems with Applications 37 (9) (2010) 6225–
6232.
30. P. Gogoi, D. K. Bhattacharyya, B. Borah, J. K. Kalita, MLH-IDS: A Multi-Level Hybrid
Intrusion Detection Method, Computer Journal 57 (4) (2014) 602–623.
31. X. Zhu, C. Jiang, L. Kuang, G. Ning, and J. Lu, “Non-orthogonal multiple access based
integrated terrestrial-satellite networks,” IEEE Journal on Selected Areas in Communications,
vol. PP, no. 99, pp. 1–1, 2017.
32. I. Guyon, A. Elisseeff, An Introduction to Variable and Feature Selection, Journal of Machine
Learning Research 3 (6) (2003) 1157–1182.
33. P. Zhang, H. Yao, M. Li, and Y. Liu, “Virtual network embedding based on modified genetic
algorithm,” Peer-to-Peer Networking and Applications, no. 2, pp. 1–12, 2017.
34. M. Ambusaidi, X. He, P. Nanda, Z. Tan, Building an Intrusion Detection System Using a
Filter-Based Feature Selection Algorithm, IEEE Transactions on Computers 65 (10) (2016)
2986–2998.
35. S. J. Horng, M. Y. Su, Y. H. Chen, T. W. Kao, R. J. Chen, J. L. Lai, C. D. Perkasa, A novel
intrusion detection system based on hierarchical clustering and support vector machines,
Expert Systems with Applications 38 (1) (2011) 306–313.
36. C. Jiang, C. Jiang, N. C. Beaulieu, L. Yong, Y. Zou, and R. Yong, “Dywamit: Asynchronous
wideband dynamic spectrum sensing and access system,” IEEE Systems Journal, vol. 11,
no. 3, pp. 1777–17882, 2017.
37. C. Elkan, Results of the KDD’99 classifier learning, ACM SIGKDD Explorations Newsletter
1 (2) (2000) 63–64.
38. Patcha, A.; Park, J.M. (2007); An overview of anomaly detection techniques: Existing
solutions and latest technological trends, Computer Networks, ISSN 1389-1286, 51(12):
3448–3470.
39. L. Kuang, C. Xi, C. Jiang, H. Zhang, and W. Sheng, “Radio resource management in future
terrestrial-satellite communication networks,” IEEE Wireless Communications, vol. 24, no. 5,
pp. 81–87, 2017.
40. J. Wang, C. Jiang, H. Zhu, R. Yong, and L. Hanzo, “Taking drones to the next level:
Cooperative distributed unmanned-aerial-vehicular networks for small and mini drones,”
IEEE Vehicular Technology Magazine, vol. 12, no. 3, pp. 73–82, 2017.
41. Lazarevic, A.; Kumar, V.; Srivastava, J. (2005); Intrusion detection: A survey, Managing
Cyber Threats, ISSN 0924-6703, 5: 19–78.
References 83

42. Axelsson, S. (1998); Research in intrusion-detection systems: a survey, Department of

Computer Engineering, Chalmers University of Technology, Goteborg. Sweden, Technical
Report 98-17.
43. H. Yao, T. Mai, X. Xu, P. Zhang, M. Li, and Y. Liu, “Networkai: An intelligent network
architecture for self-learning control strategies in software defined networks,” IEEE Internet
of Things Journal, vol. 5, no. 6, pp. 4319–4327.
44. Om, H.; Kundu, A. (2012); A hybrid system for reducing the false alarm rate of anomaly
intrusion detection system, IEEE 1st International Conference on Recent Advances in
Information Technology (RAIT), ISBN 978-1-4577-0694-3, 131–136.
45. Kaisler, S. et al (2013); Big data: Issues and challenges moving forward, IEEE 46th Hawaii
International Conference on System Sciences (HICSS), ISSN 1530-1605, 995–1004.
46. Michael, K.; Miller, K.W. (2013); Big Data: New Opportunities and New Challenges,
Computer, ISSN 0018-9162, 46(6):22–24.
47. P. Zhang, H. Yao, and Y. Liu, “Virtual network embedding based on the degree and clustering
coefficient information,” IEEE Access, vol. 4, pp. 8572–8580
48. Russom, P. et al (2011); Big Data Analytics, TDWI Best Practices Report, Fourth Quarter.
49. J. Du, E. Gelenbe, C. Jiang, H. Zhang, Y. Ren, J. Du, E. Gelenbe, C. Jiang, H. Zhang, and
Y. Ren, “Contract design for traffic offloading and resource allocation in heterogeneous ultra-
dense networks,” IEEE Journal on Selected Areas in Communications, vol. 35, no. 11, pp.
2457–2467, 2017.
50. Fan, W.; Bifet, A. (2013); Mining big data: current status, and forecast to the future, ACM
SIGKDD Explorations Newsletter, ISSN 1931-0145, 14(2): 1–5.
51. James, G. et al (2013); An introduction to statistical learning, Springer, ISSN 1431-875X.
52. H. Yao, L. Wang, X. Wang, L. Zhou, and Y. Liu, “The space-terrestrial integrated network
(stin): An overview,” IEEE Communications Magazine, vol. 56, no. 9, pp. 178–185, 2018.
53. KDD Cup 1999, http://kdd.ics.uci.edu/databases/kddcup99/kddcup99.html. May 2015
54. Lippmann, R.P. et al (2000); Evaluating intrusion detection systems: The 1998 DARPA off-
line intrusion detection evaluation, IEEE Proceedings of DARPA Information Survivability
Conference and Exposition (DISCEX), ISBN 0-7695-0490-6, 2:12–26.
55. B. Deng, C. Jiang, L. Kuang, G. Song, J. Lu, and S. Zhao, “Two-phase task scheduling in
data relay satellite systems,” IEEE Transactions on Vehicular Technology, vol. 67, no. 2, pp.
1782–1793, 2017.
56. H. Yao, C. Xu, M. Li, P. Zhang, and L. Wang, “A novel virtual network embedding algorithm
based on policy network and reinforcement learning,” Neurocomputing, vol. 284, pp. 1–9,
2018.
57. Tavallaee, M. et al (2009); A detailed analysis of the KDD CUP 99 data set, Proceedings
of the Second IEEE Symposium on Computational Intelligence for Security and Defence
Applications (CISDA), ISBN 978-1-4244-3763-4, 1–6.
58. P. Zhang, H. Yao, C. Qiu, and Y. Liu, “Virtual network embedding using node multiple metrics
based on simplified electre method,” IEEE Access, vol. 6, pp. 37 314–37 327, 2018.
59. Pfahringer, B. (2000); Winning the KDD99 classification cup: bagged boosting, ACM
SIGKDD Explorations Newsletter, ISSN 1931-0145, 1(2): 65–66.
60. Yu, G. D. et al (2014); Multi-objective rescheduling model for product collaborative design
considering disturbance, International journal of simulation modelling, ISSN 1726-4529,
13(4): 472–484.
61. Gusel, L. R. et al (2015); Genetic based approach to predicting the elongation of drawn alloy,
International journal of simulation modelling, ISSN 1726-4529, 14(1): 39–47.
62. J. Wang, C. Jiang, Z. Kai, T. Q. S. Quek, R. Yong, and L. Hanzo, “Vehicular sensing networks
in a smart city: Principles, technologies and applications,” IEEE Wireless Communications,
vol. 25, no. 1, pp. 122–132, 2018.
63. Prasad, K. et al (2016); A knowledge-based system for end mill selection, Advances in
Production Engineering & Management, ISSN 1856-6250, 11(1): 15–28.
Chapter 4
Intelligent Network Control

Finding the near-optimal control strategy is the most critical and ubiquitous problem
in a network. Examples include routing decision, load balancing, QoS-enable load
scheduling, and so on. However, the majority solutions of these problems are largely
relying on a manual process. To address this issue, in this chapter, we apply several
artificial intelligence approaches for self-learning control strategies in networks. In
this chapter, we first present an energy-aware multi-controller placement scheme
as well as a latency-aware resource management model for the SDWN. Moreover,
the particle swarm optimization (PSO) is invoked for solving the multi-controller
placement problem, and a deep reinforcement learning (DRL) algorithm aided
resource allocation strategy is conceived. Then, we present a novel controller mind
(CM) framework to implement automatic management among multiple controllers
and propose a novel Quality of Service (QoS) enabled load scheduling algorithm
based on reinforcement learning to solve the problem of complexity and pre-strategy
in the networks. In addition, we present a Wireless Local Area Networks (WLAN)
interference self-optimization method based on a Self-Organizing Feature Map
(SOM) neural network model to suppress the interference in local area networks.
Finally, we propose a BC-based consensus protocol in distributed SDIIoT, where
BC works as a trusted third party to collect and synchronize network-wide views
between different SDN controllers. In addition, we use a novel dueling deep Q-
learning approach to solve this joint problem.

4.1 Multi-Controller Optimization in SDN

Given the proliferation of mobile data traffic in a variety of wireless networks [1],
traditional network resource management techniques may not satisfy the users’
quality of experience (QoE). The software defined wireless networks (SDWN) [2]
was proposed with the spirit of decoupling the control layer and the infrastructure

© Springer Nature Switzerland AG 2019 85

H. Yao et al., Developing Networks using Artificial Intelligence,
Wireless Networks, https://doi.org/10.1007/978-3-030-15028-0_4
86 4 Intelligent Network Control

layer, which can both efficiently support big data driven resource management and
beneficially provide a global optimal system performance.
In retrospect, how to appropriately deploy controllers in different positions has
become a critical problem for wired software defined networks (SDNs), also termed
as the controller placement problem (CPP) [5], which attracts numerous heuristic
algorithms to solve it. By contrast, few works focused their attention on the CPP
in the context of SDWN architecture. Recently, Abdel et al.[6] investigated the
CPP between the controller and the elements under its control considering average
response time and maximum response time. Resource management is another key
problem, since traditional resource management methods cannot support online and
adaptive decision making corresponding to the real-world dynamic experimental
states. Hence, Grandl et al.[8] proposed a Tetris assisted resource scheduling
framework and Jalaparti et al.[10] designed a Corral aided framework, where they
defined the resource management as a combinatorial optimization problem, and
both frameworks yielded an improved performance. Moreover, given the successful
applications of deep learning algorithms in image processing and robotics, Mao
et al.[12] utilized deep reinforcement learning (DRL) algorithms for adaptively
solving resource allocation problems [25].
However, most of them only focused on the latency for multi-controller place-
ment without taking into account the impact of energy consumption imposed
on the CPP, where controllers may be energy constrained. Furthermore, resource
management of each controller in SDWNs substantially influence the QoE of users
such as the waiting time. Inspired by aforementioned issues, in this part, we propose
an energy-aware multi-controller placement scheme as well as a DRL aided resource
management model for the SDWN. The main contributions of our paper can be
summarized as follows.
• A particle swarm optimization (PSO) aided multi-controller placement model is
conceived for minimizing the system’s energy consumption with latency con-
straints. Relying on PSO, our proposed energy-aware multi-controller placement
scheme can be efficiently solved.
• Relying on the powerful expression capability of deep reinforcement learning
algorithms, we propose a DRL based resource management scheme for the
SDWN, where the self-aware controller system is capable of reducing the waiting
time of each task.

4.1.1 System Model

4.1.1.1 Network Model

In our model, the SDWN contains n controllers and m elements. Let C =

{c1 , c2 , · · · , cn } denote the set of the controllers and S = {s1 , s2 , · · · , sm } be the set
of the elements. Elements and controllers are capable of communicating with each
4.1 Multi-Controller Optimization in SDN 87

other via wireless links. And the maximum number of the elements that a single
controller can support is represented by L. Let mi be the  number of elements served
by controller ci , and we have mi ∈ (0, n − i + 1] and ni=1 mi = m. Moreover, the
distance between ci and one of its clients, i.e. si,mi , is denoted as d(ci , si,mi ). Task
j is represented by Ωj = (wj , ηj ), where wj denotes the amount of computation
of Ωj , that is the required CPU cycles for completing the task, while ηj represents
the amount of communication traffic in Ωj , i.e. the amount of the data transmitted
toward the element.

4.1.1.2 Communication Model

In our communication model, h(ci , si,mi ) represents the channel gain between
controller ci , (i = 1, 2, · · · , n, ci ∈ C) and the element si,mi served by ci . Let
p(ci ) represent the transmission power of controller ci . Hence, the uplink data rate
r(ci , si,mi ) of this link can be calculated as:
 
p(ci )h(ci , si,mi )
ri (ci , si,mi ) = B log2 1 + 2 , (4.1)
σ + I (ci , si,mi )

where σ 2 is the variance of the white Gaussian noise and B denotes the channel
bandwidth, while I (ci , si,mi ) represents the inter-cell interference between con-
troller ci and element si,mi . The transmission delay from the controller ci to the
element si,mi for task Ωj can be formulated as:

ηj
tiT (ci , si,mi ) = . (4.2)
r(ci , si,mi )

As for the transmission energy consumption of controller ci , we have:

p(ci )ηj
εT (ci , si,mi ) = p(ci )t T (ci , si,mi ) = . (4.3)
r(ci , si,mi )

4.1.1.3 Computation Model

Let us define fi as the computational capacity of the controller ci . Hence, the

controller’s execution time of task Ωj = (wj , ηj ) can be expressed as:

wj
tiC (ci , si,mi ) = . (4.4)
fi
88 4 Intelligent Network Control

Furthermore, we can obtain the execution energy consumption of controller ci for

processing task Ωj as:

εiC (ci , si,mi ) = ρi wj , (4.5)

where ρi is the unit execution energy consumption of one CPU cycle in the
controller.

4.1.1.4 Problem Formulation

Considering both the limited computational capacity and the energy constraint of
controllers, we propose an energy-aware multi-controller placement scheme as well
as a latency-aware resource management model for the SDWN. Our objective is to
minimize the average energy consumption of all controllers for both communication
and computation in large-scale SDWNs. This placement problem can be formulated
as follows:

1  T
n
min εi (ci , si,mi ) + εiC (ci , si,mi )
n
i=1

s.t. (4.6-a) : tiT (ci , si,mi ) + tiC (ci , si,mi ) ≤ tmax , ci ∈ C, si,mi ∈ S,
(4.6-b) : d(ci , si,mi ) ≤ dmax , ci ∈ C, si,mi ∈ S, (4.6)
(4.6-c) : mi ≤ L(ci ), ci ∈ C

n
(4.6-d) : mi = m,
i=1

where tmax in (4.6-a) represents the maximum time requirement of the task, while
dmax in (4.6-b) is the maximum deployment distance between ci and si,mi . L(ci )
in (4.6-c) represents the maximum number of the elements that the controller ci
can support. In Sect. 4.1.2, PSO algorithm will be invoked for solving our proposed
multi-controller placement problem.
After obtaining a delicate multi-controller placement strategy from (4.6), we try
to further determine computational resource allocation scheme of the controller for
each task. Hereinafter, DRL algorithm is employed to solve this problem. For the
task Ωj , its ideal completion time is defined by Tideal,Ωj , while the actual execution
time is defined by Tactual,Ωj . Hence, we aims for minimizing the its waiting time
ΓΩj , i.e.

min ΓΩj = Tactual,Ωj − Tideal,Ωj . (4.7)

4.1 Multi-Controller Optimization in SDN 89

4.1.2 Methodology
4.1.2.1 PSO Aided Near-Optimal Multi-Controller Placement

The PSO is a family member of stochastic optimization algorithms, which is

inspired by the social behavior of birds [45]. In PSO, for the i-th particle
in the d-dimensional solution space, its position can be represented as Xi =
(xi1 , xi2 , · · · , xid ) and its velocity can be represented as Vi = (vi1 , vi2 , · · · , vid ).
Relying on the fitness function, the individual best position of particle i
is expressed as Pi = (pi1 , pi2 , · · · , pid ), while its global best position is
Pg = (pg1 , pg2 , · · · , pgd ). Hence, the velocity in the iteration (n + 1) can be
updated by:

Vi (n + 1) = ωVi (n) + c1 ξ1 (Pi − Xi ) + c2 ξ2 (Pg − Xi ), (4.8)

where c1 and c2 are two positive constants generated in [0, 2], while ξ1 and ξ2 are
a pair of independent random numbers valued in (0, 2). ω is a weight coefficient
for the sake of ensuring the convergence of PSO algorithm. Based on the updated
velocity, particle’s position can be given by:

Xi (n + 1) = Xi (n) + Vi (n + 1). (4.9)

In this part, we apply PSO algorithm to solve the multi-controller placement

problem, where particles are represented as the possible position of the controllers.
After iterations, a near-optimal solution of controller placement can be obtained, as
shown in Algorithm 1.

4.1.2.2 Resource Management Relying on Deep Q-Learning

In the following, we study the computational resource allocation scheme of the

controller for each task relying on DRL algorithm [4, 11]. In the process of
interacting with the environment, the model-free Q-learning algorithm is capable
of finding the near-optimal behavior by continuous trial and error. Assume the state
in the step n is presented as Sn , π means the policy space, and the accumulated
reward V π (Sn ) of state Sn is calculated as:

V π (Sn ) = Rn + γ Rn+1 + γ 2 Rn+2 + · · · , (4.10)

where Rn is the instant reward in step n, while γ (0 < γ < 1) represents the
discount factor which reflects the influence of the future reward. More explicitly,
a large γ means that the training agent focuses more on to the future experience,
otherwise the training agent only concerns about the immediate reward [55]. In Q-
learning algorithm, the Q-value is the function of state S and action A, which is
formulated as:

Q(Sn , An ) = Rn + γ V π (Sn+1 ). (4.11)

90 4 Intelligent Network Control

Algorithm 1 PSO aided multi-controller placement

Input: network topology, objective function, the number of controllers m, the number of
particles s;
Output: best controller placement x ∗ ;
Initialization:
Randomly generate particle’s position Xi and velocity Vi ;
Calculate the fitness value of each particle according to objective function;
Obtain the individual best position Pi and the global best position Pg .

while i < n do
for each particle Xi
Update particle’s velocity Vi according to Eq. (4.8);
Update particle’s position Xi according to Eq. (4.9);
Calculate the fitness value of particle F itness(Xi );

if F itness(Xi ) < F itness(Pi ) then

Update the individual best position Pi ;

end if
if F itness(Xi ) < F itness(Pg ) then
Update global best position Pg ;

end if
end while

Thus, Qn+1 (Sn , An ) can be updated by:

Qn+1 (Sn , An ) = (1 − η)Qn (Sn , An )

(4.12)
+ η[Rn + γ max Qn (Sn+1 , An+1 )],

where the learning rate 0 < η < 1 controls the learning convergence speed.
Deep Q-network utilizes a feedforward artificial neural network for the sake
of approximating the Q-value function, i.e Q(Sn , An ; θ ). The deep Q-network is
trained in the form of minimizing a loss function by updating θ in terms of small
steps. The loss function can be given by:

L(θ ) = E[(y(Sn , An , Sn+1 ; θ̂ ) − Q(Sn , An ; θ ))2 ], (4.13)

where the target function y(Sn , An , Sn+1 ; θ̂ ) = R + γ max Q(Sn+1 , An+1 ; θ̂ ).

In this part, we apply the deep Q-network algorithm to solving the resource
management problem, as shown in Algorithm 2. For each SDWN controller, we
consider there are total Ω tasks waiting for service. Three key factors of the deep
Q-network are given as follows.
• State Space: The available resources of controllers includes the current allocated
computational resources and the required computational resources of the tasks
which may be scheduled [62]. We define the demand of computational resources
4.1 Multi-Controller Optimization in SDN 91

of the task j with type z by uj,z , and the current computational resource level
of the controller i as ui,z . Hence, considering a total of Z kinds of tasks, the
system’s state with total m controllers can be given by Λ = [S1 , S2 , · · · , Sm ],
where Sm = [uj,k , um,k , z = 1, 2, · · · , Z].
• Action Space: The serving queue can hold maximum K tasks and each controller
is capable of dealing with only one task in each time slot. For the controller m,
let the action Am = k represent the k-th task is scheduled at the current time slot,
while Am = φ means that no task is scheduled in the current time slot. Then the
action space of controller m can be expressed as {φ, 1, 2, · · · , K}.
• Reward: The reward function is designed for directing the agent to minimize the
average waiting
 time in the controllers. Specifically, we set the reward at each
time slot to 1/Tideal,Ωj , Ωj ∈ Ω, where Ω is the set of current tasks supported
by the controller.

Algorithm 2 Deep Q-network based resource management strategy

Input: state S, action A and the maximum number of iterations N
Output: Q(S, A);
Initialization:
Initialize the weight of deep Q-network θ;

for n < N do
if random probability p < δ then
select an action An ;
otherwise
An = arg max Q(Sn , An ; θ)
end if
Execute action An , then obtain the reward Rn and arrive at the next state Sn+1 ;
Calculate the target Q-value
y(Sn , An , Sn+1 ; θ̂) = R + γ max Q(Sn+1 , An+1 ; θ̂)
Update the deep Q-network by minimizing the loss L(θ);
L(θ) = E[(y(Sn , An , Sn+1 ; θ̂) − Q(Sn , An ; θ))2 ]
end for

4.1.3 Simulation Results

In this subsection, we evaluate proposed PSO algorithm relying on six different

network topologies from the Internet zoo [7, 14]. The channel bandwidth is B =
5 MHz and the transmission power is p = 50 mW. The variance of white Gaussian
noise is σ 2 = −100 dBm. The channel gain is set as h = d −λ , where λ = 4
represents the path loss factor. The amount of one task is 20 Mb. ρ = 4 J/GHz
denotes the unit energy consumption per CPU circle. The computational capacity of
the controller is f = 2 GHz. In order to verify the performance of our proposed
algorithm, we use the genetic algorithm (GA) as well as the greedy algorithm
(GL) for comparison. The number of particles is 20. Moreover, some parameters
of proposed two algorithms are listed in Table 4.1.
92 4 Intelligent Network Control

Table 4.1 Parameters of two Algorithm Parameters

algorithms
PSO c1 = 2.0, c2 = 2.0, w = [0.4, 0.9]
GA Pc = 0.8, Pm = 0.1

a 340 b 380
PSO PSO
GA GA
Energy consumption

Energy consumption
330 GL 370 GL

320 360

310 350

300 340
0 10 20 30 40 50 0 10 20 30 40 50
Iterations Iterations
c d
360 440
PSO PSO
GA GA
Energy consumption
Energy consumption

350 GL 430 GL

340 420

330 410

320 400
0 10 20 30 40 50 0 10 20 30 40 50
Iterations Iterations
e 340 f 390
PSO PSO
GA 385 GA
Energy consumption

Energy consumption

330 GL GL
380

320 375

370
310
365

300 360
0 10 20 30 40 50 0 10 20 30 40 50
Iterations Iterations

Fig. 4.1 Energy consumption versus iterations in six different network topologies (a) Aarnet. (b)
Arn. (c) Bandcon. (d) Bellcanada. (e) Esnet. (f) ChinaNet

Figure 4.1 shows the performance of energy consumption versus iterations in six
different topologies which assumed the link between nodes is wireless, i.e. Aarnet,
Arn, Bandcon, Bellcanada, Esnet and ChinaNet. As shown in Fig. 4.1, with the
4.1 Multi-Controller Optimization in SDN 93

increase of the iteration, PSO has the minimum energy consumption in comparison
to the other two algorithms in the context of m = 4 controllers. We can conclude that
the proposed PSO algorithm can provide a beneficial position placement scheme
having the minimum energy consumption considering the latency constraints. By
contrast, GL has the worst performance because it is easy to be trapped into local
optimum. Hence, simulation results show that the PSO algorithm can obtain the
near-optimal solution of controller placement for SDWNs.
As for the resource allocation, we use a neural network having a fully connected
hidden layer with 30 neurons to construct our deep Q-learning network. The
network’s parameters are updated with learning rate 0.002. There are total three
kinds of tasks, and the serving queue can hold at most 20 tasks. Moreover, the
ideal completion time of each task is randomly generated from [1,10]. The
simulation results are shown in Fig. 4.2. In Fig. 4.2a, we show the average waiting
time versus the controller’s load compare with both the short job first (SJF)
algorithm and random serving scheme. We can conclude that with the increase of
controller’s load, our proposed DRL algorithm outperforms the others, which can
improve the efficiency of the controllers. Moreover, as shown in Fig. 4.2b, a large

10 8
DRL
Average waiting time

SJF
Average waiting time

8 RANDOM 6

6 4

4 2

2 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Load of the controller Possion arrival rate

8 25
Average waiting time

Average waiting time

20
6
15
4
10
2
5

0 0
2 5 10 15 20 25 30 35 5 10 15 20 25 30 35 40 45 50
Resources in the controller Tasks in the queue

Fig. 4.2 Average waiting time versus different parameters. (a) Load in the controller. (b) Task
arrival rate. (c) Resources in the controller. (d) Queue length
94 4 Intelligent Network Control

Poisson arrival rate of tasks process increases the average waiting time of the system.
In Fig. 4.2c, it can be seen that the average waiting time of the system is reduced
with the increase of computational resources of the controllers. Finally, Fig. 4.2d
shows that the number of tasks in the queue increases the average waiting time of
the system.

4.2 QoS-Enabled Load Scheduling Based on Reinforcement

Learning

Energe resources crisis and global warming have become two global concerns [92].
As reasonable solutions, smart grid [16] and Energy Internet (EI) are seen as a
new generation of energy provision paradigm, where improved communication
mechanism is important to enable end-to-end communication. Software-defined
networking (SDN) [18] is seen as a promising paradigm shift to reshape future
network architecture, as well as smart grid and EI, called software-defined EI
(SDEI). Using SDN enables to improve smart grid and EI by providing an
abstraction of underlying network resources, forming global view for applications
from upper layers, and decoupling infrastructures and control plane to enhance
the flexibility and reliability of the system [21]. Noteworthy, the control plane is
considered as the brain of SDN [22]. With the explosion of network scales and
network traffic, overload in a single controller is one of the most intractable issues
[118]. There is a growing consensus that the control plane should be designed
as a multiple controllers plane to constitute a logically centralized but physically
distributed model [27, 28, 32]. So far, the issues of multiple controllers have been
studied in literature. Except for addressing the consistency problem of global view
among distributed control plane, another key issue is how to schedule loads among
multiple controllers so as to mitigate the risk of overloads and failures in one single
controller.
On the other hand, the most important application of SDN in smart grid is
real-time monitoring and communicating [96]. It follows that these applications
require steady web-environment with no packet loss and less time delay to keep
high accuracy and real time capability [33].
Traditionally, load scheduling algorithms make load scheduling decisions after
the overload problems have happened [102]. In general, the traditional algorithms
have three steps, including collecting load information, making load scheduling
decisions, and sending load scheduling commands to the corresponding controllers.
For example, the work in [34], load scheduling decision is made after the problem
of overload. In addition, current CPU usage, current memory usage, current hard
disk usage, and weight coefficient need to be exchanged among controllers when
new load scheduling decision is made, which occupies lots of extra time so as to
decrease time efficiency.
4.2 QoS-Enabled Load Scheduling Based on Reinforcement Learning 95

Recently, Machine learning (ML) has emerged as a novel technique, which

can be used to tackle the above challenges [106]. Reinforcement Learning (RL)
is regarded as an important branch of ML to solve complex control problems.
RL is quite different from the traditional management and operation methods. It
develops computer models to train datasets, which resolves the problems without
being explicitly programmed to, but learning from environments.

4.2.1 System Description

In this subsection, we first give brief overviews of energy Internet and software-
defined energy Internet. Then the CM framework in SDEI is presented.

4.2.1.1 Energy Internet

With energy crisis and limitation around world, how to use renewable energy has
attracted lots of attentions, ranging from government and industry to academia [15].
Here, the development of renewable energy, as well as information and communica-
tion technologies (ICTs) are two key enablers of energy Internet [13, 19, 50]. Thus,
energy Internet can be seen as an energy-utilizing system, combining distributed
renewable energy with the advanced ICTs, which is known as the version of smart
grids 2.0 [52].
Specially, ICTs provide a viable way to use the control capability of smart grid
and allow distributed energy to access to the backbone grid in EI [9, 53]. Here, smart
grid is used to collect and operate the information about the behaviors of users and
suppliers to improve the sustainability and reliability of energy.

4.2.1.2 Software-Defined Energy Internet

With traditional TCP/IP protocol, energy Internet has achieved great success [111].
However, many challenges have emerged with the increasing number of smart
connected devices in smart grid. It is hard for such rigid and static Internet to meet
the demands of flexibility, agility, and ubiquitous accessibility.
In order to solve this embarrassment, there is a consensus to establish the future
energy Internet architecture. SDN is seen as one of the most promising paradigms
[54]. It is an approach to implement the network that separates the control plane and
the data plane, abstracts the underlying infrastructures, and simplifies the network
management by introducing the ability of programming [116].
96 4 Intelligent Network Control

Some works have employed SDN in energy Internet and smart grid [23]. For
example, in order to support secure communications, the authors in [56] learned a
SDN-enabled multi-attribute secure architecture for smart grid in IIoT environment.
Moreover, the authors in [57] proposed a software defined advanced metering
infrastructure (AMI) communication architecture. By this architecture, the problem
of global load-balanced routing was solved.
Based on these works, we consider a software-defined energy Internet. However,
before the wide adoption of SDEI, there are some problems remaining to be solved.
The most intractable one is the scalability and reliability of control plane in SDEI.
It can be anticipated that a logically centralized, but physically distributed control
plane is necessary. Thus, we propose a controller mind framework in distributed
SDEI to implement automatic management among multiple controllers.

4.2.1.3 Controller Mind framework

Based on traditional SDN architecture, we propose CM framework bridging the

control plane and the data plane transparently, as shown in Fig. 4.3.

Fig. 4.3 A framework of controller mind in SDEI

4.2 QoS-Enabled Load Scheduling Based on Reinforcement Learning 97

Fig. 4.4 The detailed CM framework of QoS-enabled load scheduling

The detailed structure of CM framework is given in Fig. 4.4. CM takes the

responsibility of re-queuing incoming Packet-in flows to guarantee QoS and for-
warding them to the appropriate controllers based on the results of reinforcement
learning. Therefore, the CM framework consists of three modules, including re-
queuing module, info-base module and learning module.

4.2.1.4 Re-Queuing Module

We assume that there are two types of traffic flows, namely QoS flows with high-
priority and best-effort flows with low-priority. QoS flows include some traffic about
the application of real-time monitoring in SDEI, and best-effort flows include some
traffic from other applications with the low-level requirement of real-time capability
[24]. When these traffic flows are encapsulated into Packet-in messages and sent to
CM, based on source/destination MAC address, IP address, and TCP/UDP port in
the packet headers, the re-queuing module marks and classifies the incoming Packet-
in messages as QoS flows and best-effort flows [29, 35, 59], then re-queues them by
the method shown in Sect. 4.2.2.1.

4.2.1.5 Info-Table Module

It receives re-queuing Packet-in messages from re-queuing module, sends them

to controllers based on the learning results from the learning module, receives
Flow-mod messages from the control plane, and sends Flow-mod messages to the
data plane. The corresponding changes of the loads in each controller are aware
by sending Packet-in messages, and receiving Flow-mod messages, which are all
recorded by this module. On the one hand, these load records are the training
98 4 Intelligent Network Control

datasets of learning module, on the other hand, by this mechanism, the frequent
signaling interactions that are used to obtain the current load information are
avoided, compared with the traditional schemes.

4.2.1.6 Learning Module

Based on the historical load records from the info-table module and reinforcement
learning algorithm, learning module trains the data offline, obtains the learning
results, and sends to info-table module. The reinforcement learning algorithm, i.e.,
Q-learning, is executed in this module, and the detail of the algorithm will be shown
in Sect. 4.2.3.

4.2.2 System Model

In this subsection, we present re-queuing model, followed by workload model.

4.2.2.1 Re-Queuing Model

If we use FIFO (First In First Out) model, some delay-sensitive flows can’t
be treated fairly. Therefore, when arriving at CM, Packet-in messages would be
classified into QoS messages or best-effort messages by extracting their headers,
such as source/destination MAC address, IP address and TCP/UDP port. The
architecture of classification and re-queuing at CM is shown in Fig. 4.5.
Here, we use a weight fair queuing(WFQ)[20, 60] algorithm in re-queuing model.
In WFQ, we give weight coefficient wi to classify queue i. Each classified queue
sends the messages based on weight coefficient wi . When the queue is empty, skip
and access the next queue. Hence, the average service time of queue i is wwi , where
 j
wj is the sum of weight coefficients of all non-empty queues.

Fig. 4.5 Classification and re-queuing at CM

4.2 QoS-Enabled Load Scheduling Based on Reinforcement Learning 99

4.2.2.2 Workload Model

With the rapid development of commercial deployment, the performance of SDN

controller is more and more important [36]. Global SDN certified testing cen-
ter (SDNCTC) [61] develops a SDN controller performance test tool, called
OFsuite_Performance, and releases a RYU controller performance test report
[17, 63]. In this report, when the arriving rate of Packet-in messages is lower,
the flow response time is linear with the Packet-in messages arriving rate, and the
acceptable maximum arriving rate of Packet-in messages is related to the controller’s
own performance parameters. Hence, from the test result in the report, we obtain
the relationship between the response time and the number of Packet-in messages in
(4.14),

τ = ρNp + β, Np ≤ max, (4.14)

where τ represents the response time, and Np is the number of Packet-in messages,
ρ and β are the parameters related to the performance of each controller. max is
the maximum number of Packet-in messages in the controller. This equation is
an empirical expression that is shown in RYU controller performance test report.
Although it has not ever been used in previously published work, we can have this
relationship according to the real test works.
Meanwhile, Zhang et al. in [65] provided the relationship between the response
time and the servers’ load status as shown in (4.15),

τ = θ ls , (4.15)

where ls is the load status of servers. θ is a parameter related to server’s perfor-

mance. Controllers are always deployed in servers, so in this part we use the same
load status model in controllers as servers.
From (4.14) and (4.15), we can deduce (4.16), which explains the relationship
between the load status and the number of Packet-in messages in the controller.

ls = logθ (ρNp + β), Np ≤ max. (4.16)

When Np = max, the load status is 100%. Thus using (4.16), we have
logθ (ρmax + β) = 1, and there is a necessary relationship between the parameters,
which is ρmax + β = θ . Obviously, when Np > max, the load status is also 100%.
Thus we have (4.17),

logθ (ρNp + β) Np ≤ max
ls = , (4.17)
100% Np > max

where ρmax + β = θ
100 4 Intelligent Network Control

In the following problem formulation and simulation, we use (4.17) as the

workload model. Since the number of Packet-in messages Np is recorded by the
info-table module, we use (4.17) to transfer the number of Packet-in messages to
the load status of controllers to formulate and simulate the problem in the following
sections.

4.2.3 Problem Formulation

In this subsection, we first describe a brief review of reinforcement learning. Then

we formulate the load scheduling problem as a Q-learning process, which starts with
the optimization problem formulation.

4.2.3.1 Reinforcement Learning

Figure 4.6 shows the agent-environment interaction model of reinforcement learn-

ing. At the decision epoch, the agent obtains environment state s and corresponding
reward r. According to the given policy, the agent selects action a to work on the
current environment, which makes state s turn into new state s . Then at the next
decision epoch, the agent and environment interact in the same way.
The agent selects the action based on the policy function which defines the behav-
ior at a given moment. A stationary random policy is defined as π : S × A → [0, 1],

Fig. 4.6 The agent-environment interaction model of reinforcement learning

4.2 QoS-Enabled Load Scheduling Based on Reinforcement Learning 101

where π [s, a] represents the probability of selecting action a under state s, S is state
space, and A is action space.
There are two value functions to represent the feedbacks from each decision,
namely state value function V π (s) and action-state value function Qπ (s, a). V π (s)
means the expected total rewards based on policy π in state s, and it can be
represented as:
∞

V π (s) = E π [ γ k rt+k+1 |st = s], (4.18)
k=0

where E π [∗] denotes the mathematical expectation under state transferring proba-
bility P (s, a, s ) and policy π . rt is immediate reward at time t. γ ∈ (0, 1] denotes
the discount factor to trade-off the importance of immediate reward and long-term
reward.
Additionally, action-state value function Qπ (s, a) represents the expected total
rewards based on policy π in state-action pair (s,a), and it can be represented as:
∞

Qπ (s, a) = E π [ γ k rt+k+1 |st = s, at = a]. (4.19)
t=0

And there is a relationship between V π (s) and Qπ (s, a).

For a certain policy π ,
V π (s)= Qπ (s, π(s)). For a stochastic policy, V π (s) = a∈A π(s, a)Qπ (s, a).
Hence, Qπ (s, a) can be expressed by V π (s) as follows:

Qπ (s, a) = R(s, a) + γ P (s, a, s )V π (s ), (4.20)
s ∈S

where R(s, a) = s ∈S P (s, a, s )R(s, a, s ) means the expected reward of

selecting action a under state s. R(s, a, s ) = E[rt+1 + γ rt+2 + γ 2 rt+3 + . . . |st =
s, a = at = π(st ), st+1 = s ] denotes the expected reward of selecting action a to
transfer the state from s to s .

4.2.3.2 Reinforcement Learning Formulation

In order to obtain the optimal policy, it is necessary to define state space, action
space and reward function in Q-learning model. Before this, we will give the
optimization problem formulation in the RL based QoS-enabled load scheduling
problem.
102 4 Intelligent Network Control

4.2.3.3 Optimization Problem Formulation

Our target is to find out the optimal scheme to allocate Packet-in messages from the
data plane to the control plane with the minimum waiting time of QoS flows and the
acceptable packet loss rate of best-effort flows. The minimization problem can be
formulated as the weighted sum as follows.


T 
N1 
T 
N2
min k1 T Q1i (t) + k2 P LQ2k (t), (4.21)
t=0 i=1 t=0 k=1

subj ect to

P LQ1i (t) = 0, ∀i = 1, 2, . . . N1 , (4.22)

where we assume there are T time slots during the whole system, which starts when
the first Packet-in message comes and terminates when the last Packet-in message
departs. Let t ∈ {0, 1, 2, . . . , T − 1} denote the time instant. T Q1i (t) is the waiting
time of QoS flows i at time instant t. P LQ1i (t) and P LQ2k (t) are the packet loss
rates of QoS flows i and best-effort flows k at time instant t, respectively. N1 and
N2 are the total number of QoS flows and best-effort flows, respectively. k1 and k2
are the scale factors, and k1 + k2 = 1.
In the above optimization problem, the constraint (4.22) guarantees that the QoS
flows have no packet loss.
Notably, one of the optimal target in (4.21) is to minimize the packet loss rate
of best-effort flows, which is equivalently substituted by the load variation among
all controllers in the remainder of this section. Because best-effort messages have
the low priority, i.e., when the messages loss happens, it is probable that best-effort
messages are discarded. Thus, the lower load variation leads to the lower packet loss
rate of best-effort messages directly.

4.2.3.4 State Space

In order to reduce the load variation among all controllers and the waiting time of
QoS flows, we propose the definition of state space as follows:

S = {s = [Qincom , lc , qc ], |Qincom ∈ Qlevel , lc ∈ Lc , qc ∈ Qc }, (4.23)

where

Qlevel = {1, 2}, (4.24)

Lc = {[lc1 , lc2 , · · · , lck , · · · , lcN ]}, (4.25)
Qc = {[Qc1 1 , Qc2 1 , · · · , Qck 1 , · · · , QcN 1 ]}, (4.26)
4.2 QoS-Enabled Load Scheduling Based on Reinforcement Learning 103

where N is the total number of controllers in the system, and ck means the kth
controller. Qlevel means the different QoS levels of flows in this system. When
Qlevel = 1, this flow is QoS flow with the high-priority. When Qlevel = 2, this
flow is best-effort flow with the low-priority. Lc means the set of the load status of
all controllers, and lck denotes the load status of controller ck which is calculated by
(4.17) and the number of Packet-in messages is recorded by the info-table module.
Qc means the set of the number of QoS flows in all controllers, and Qck 1 denotes the
number of QoS flows in controller ck , which is recorded by the info-table module.

4.2.3.5 Action Space

In the system, the agent has to decide how to allocate Packet-in message among
multiple controllers. Thus, the action space A of RL can be defined as follows,

A = {ac1 , ac2 , · · · , ack , · · · , acN }, (4.27)

where ack represents the allocation control between the current Packet-in message
and controller ck . if ack = 1, the current Packet-in message is assigned to controller
ck . if ack = 0, the current Packet-in message is not assigned to the controller ck .

Note that N k=1 ack = 1, which guarantees that the current Packet-in message only
has one assigned controller.

4.2.3.6 Reward Function

We define numerical reward r that the agent obtains from taking action a at state s.
We have two targets as shown in (4.21), including minimizing the load variation and
the waiting time of QoS flows. Accordingly, there are two parts in reward function
r, including the standard deviation of all controllers and the number of messages
whose QoS levels exceed the incoming message, respectively.
The lower standard deviation means the better load balancing. Since bigger
reward is taken in Q-learning, we use the negative standard deviation to represent
the load variation in reward function, which is denoted in the first part of reward
function r.
Since all controllers in the system are QoS-enabled, which means Packet-in
messages will re-queue after arriving at all controllers to make sure QoS flows to be
processed with the high priority. Thus, the waiting time of incoming QoS flows is
only related to the number of QoS flows before them. The fewer QoS flows lead to
the less waiting time, which is denoted in the second part of reward function r.
In summary, reward function r can be expressed as follows:

r = − k2 std(lc1 , lc2 , · · · , lcN )

N

ck (4.28)
− k1 bool(Qincom ≥ Qqueue (i)),
i=1
104 4 Intelligent Network Control

where Nck is the number of Packet-in messages in controller ck . Qincom represents

the QoS level of incoming message, and Qqueue (i) means the QoS level of the ith
messages in controller ck . bool denotes the Boolean operator. For example, when
the QoS level of the incoming message is 1 and the current messages’ QoS line
 Nc
of one controller is 11222, the result of i=1k bool(Qincom ≥ Qqueue (i)) is 2. In
the other controller, its current messages’ QoS line is 11122, the result of boolean
operation is 3. If only consider the waiting time of incoming message, the agent
is more possible to allocate this message to the first controller because of the less
waiting time. k1 and k2 are the scale factors which are the same as (4.21).

4.2.3.7 Method to Learn the Optimal Strategy

Q-learning is a typical algorithm of reinforcement learning, and we use Q-learning

to learn the optimal strategy in this part, where action-state value function Q(s, a) is
selected as the estimation function, rather than state value function V (s). The basic
idea of Q-learning is to evaluate Q(s, a) by a temporal difference method, which is
denoted as follows:

Q(st , at ) ← Q(st , at ) + α(rt + γ max Q(st+1 , a) − Q(st , at )), (4.29)

a

where α denotes the learning efficiency. In Q-learning, each Q(s, a) is put into Q-
table.
At first, Q-learning initializes the Q−table. Then at state st , the agent determines
action at according to ε-greedy policy, and obtains the experience knowledge as
well as the training samples (st , at , st+1 , at+1 ). Meanwhile, the agent uses (4.29)
to update Q(st , at ) and Q−table. When meeting with the goal state, the agent
terminates one loop iteration. Then Q-learning continues a new loop iteration from
the initial state until the end of learning. The algorithm performed on each step is
shown in Algorithm 1.

Algorithm 3 Q-learning
1: Initialize Q−table, and set parameters α, γ and k0 ;
2: for k = 1 : k0 do
3: Select a initial state st randomly
4: while st ! = sgoal do
5: Select action at based on ε-greedy policy, and obtain immediate reward rt and next
state st+1
6: Q(st , at ) ← Q(st , at ) + (rt + γ max Q(st+1 , a) − Q(st , at ))
a
7: st ← st+1
8: end while
9: end for
4.2 QoS-Enabled Load Scheduling Based on Reinforcement Learning 105

4.2.4 Simulation Results and Discussions

In this subsection, we use computer simulation to evaluate the performance of RL

based QoS-enabled load scheduling. Firstly, we describe simulation settings, then
present the simulation results. In this subsection, we use computer simulation to
evaluate the performance of RL based QoS-enabled load scheduling. Firstly, we
describe simulation settings, then present the simulation results.

4.2.4.1 Network Topology

We choose the same topology as the one in [34], which has three controllers in the
control plane, and several switches in the data plane. Thus, N = 3.

4.2.4.2 Parameter Settings

The different seeds are employed in the simulation, and performances are average to
estimate the performance of our proposed scheme. We utilize the queuing theory to
model the arrival, processing and departure of Packet-in messages. Here, the arrival
of the Packet-in messages is based on a Poisson distribution with the parameter of
λ, indicated by the arriving rate of Packet-in messages. The processing time of each
controller is based on the negative exponential distribution with the parameter of μ,
indicated by the performance of controllers. And we assume that all controllers have
the same performance, i.e., the same μ. Summarily, the values of all parameters in
the simulation are summarized in Table 4.2.

Table 4.2 Parameters setting in the simulation

Parameter Value Description
max 20 The maximum number of Packet-in messages in controllers
ρ 0.5 The performance parameter of the controller
β 5 The performance parameter of the controller
θ 15 The performance parameter of the controller
N 3 The number of total controllers in the control plane
k1 0.6 One scale factor
k2 0.4 Another scale factor
ε 0.9 The greed factor
γ 0.65 The discount factor
Initialized α 1 The learning efficiency
μ 16 Service rate of each controller
106 4 Intelligent Network Control

For performance comparison, four schemes are simulated:

• RL based QoS-enabled load scheduling scheme proposed in this part and we call
it as RL in the remainder of this section.
• Dynamic weight based QoS-enabled load scheduling scheme in the work of [34]
and we call it as DW in the remainder of this section.
• QoS-enabled scheme, and this scheme does not take consideration of the load
scheduling. We call it as QS in the remainder of this section.
• Mini-connected load scheduling scheme, and this scheme does not consider the
QoS. We call it as MN in the remainder of this section.

4.2.4.3 Performance Evaluation Results

Figure 4.7 shows the relationship between the load variation and the Packet-in
messages arriving rates of different schemes when the proportion of QoS messages
is 75%. With the increase of the arriving rate, the load variation is increasing.
The reason is that, as the arriving rates increasing, more messages accumulate in
controllers, which obviously results in the larger load variation. In any case of
arriving rate, QS’s load variation is much bigger than others’, because QS scheme
only considers the priority of messages and fails to take the load balancing into
consideration, some controllers are overloaded and others are idle, which leads to
the biggest load variation. Taking the load balancing into consideration, the other
three schemes’ load variations are much smaller. Relatively speaking, DW’s load
variation is bigger. The reason is that, the adjustment of the load does not happen

2.5
RL based Qos-enabled load scheduling
dynamic weight based QoS-enabled load scheduling
Qos-enabled scheme
2 miniConnect load scheduling
Load variation

1.5

0.5

0
2 4 6 8 10 12 14
Packet-in messages arriving rate (packet/s)

Fig. 4.7 Load variation versus Packet-in messages arriving rates of RL, DM, QS and MN
4.2 QoS-Enabled Load Scheduling Based on Reinforcement Learning 107

at each step in DW. Only when overloaded, the dynamic weight load balancing is
triggered. But in MN, when each message arrives, it is assigned to the controller with
the least load status, which is equivalent to adjust the load distribution scheme in
each step. So MN is better than DM. But in any case of the arriving rates, RL’s load
variation is very close to the MN’s curve. Even the RL’s load variation is smaller
than the MN’s in some cases. Because by the offline learning of the historical data,
RL performs the optimal load distribution globally, which results in the best load
scheduling effect.
Figure 4.8 displays the relationship between the waiting time of QoS messages
and Packet-in messages arriving rates of different schemes when the proportion
of QoS messages is 75%. For QS scheme, although it considers the priority of
messages to let the messages with the high priority be processed firstly, no load
balancing mechanism also results in more waiting time of QoS messages. In the
low arriving rates case, DW’s waiting time is less than MN’s. The reason is that,
when the arriving rate is lower, it is unlikely to be overloaded in controllers, but MN
scheme needs to get the load status of controllers by exchanging the signaling to
adjust the load distribution in each step, which results in the additional time delay.
DW scheme isn’t triggered in the low load status, so it does not lead to the time
delay of the signaling exchange and has relatively smaller time delay, compared
with the MN scheme. With the increase of arriving rates and messages accumulating
in controllers, MN and DW schemes also exchange the signaling frequently, but
MN has the better load balancing performance, as shown in Fig. 4.7, it also has
the better time efficiency, compared with DW. And for RL scheme, because the

1.4
RL based Qos-enabled load balancing
Average waiting time of QoS messages(ms)

Dynamic weight
1.2 Qos-enabled scheme
miniConnect load balancing

0.8

0.6

0.4

0.2

0
2 4 6 8 10 12 14
Packet-in messages arriving rate (packet/s)

Fig. 4.8 Average waiting time of QoS messages versus Packet-in messages arriving rates of RL,
DM, QS and MN
108 4 Intelligent Network Control

0.8

0.7 RL based Qos-enabled load scheduling

dynamic weight based QoS-enabled load scheduling
Qos-enabled scheme
0.6 miniConnect load scheduling
Load variation

0.5

0.4

0.3

0.2

0.1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
The proportion of Qos messages

Fig. 4.9 Load variation versus the proportion of QoS messages in RL, DM, QS and MN

allocated scheme has been learned offline and in advance, and it is no need for RL
scheme to exchange the signaling at all. So in the lower arriving rates, RL scheme
has no additional time delay. In the higher arriving rates, RL scheme has a little time
delay because of the increasing of messages. Overall, RL scheme has the best time
efficiency.
Figure 4.9 presents the load variation when the proportion of QoS messages
changes at the arriving rate of 8 packet/s. Because the arrival rate is constant, the
load variation has no relationship with the proportion of QoS messages. But we can
draw the similar conclusion as Fig. 4.7, which is that RL scheme has the best load
variation.
Figure 4.10 shows the relationship between the waiting time of QoS messages
and the proportion of QoS messages at the arriving rate of 8 packet/s. For QS, with
the growth of QoS messages, the waiting time increases linearly, because it only
considers the priority of messages but no load balancing. In the lower proportion
case, DW’s waiting time is less than MN’s. The reason is that, only when overload
happens, DW scheme is triggered. And MN scheme happens in each decision epoch.
Under the current arriving rate, it is unlikely to be overloaded. So DW has relatively
smaller time delay, compared with MN in the lower proportion. The increase of
proportion leads to the growth of time delay directly. MN has the better load
balancing as shown in Fig. 4.9, which results in the better time efficiency in the
higher proportion. RL enables to learn the allocated scheme in advance and offline
with no signaling exchanging. So when QoS messages are smaller, RL has no time
delay completely. And with the growth of QoS messages, RL has a litter time delay
and the best time efficiency.
4.3 WLAN Interference Self-Optimization Based SOM Neural Networks 109

0.6

Average waiting time of QoS messages(ms)

RL based Qos-enabled load balancing
Dynamic weight
0.5 Qos-enabled scheme
miniConnect load balancing

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
The proportion of Qos messages

Fig. 4.10 Average waiting time of QoS messages versus the proportion of QoS messages in RL,
DM, QS and MN

4.3 WLAN Interference Self-Optimization Based SOM

Neural Networks

An important difference between WLAN (Wireless Local Area Networks) and other
commercial mobile communication networks like 3G, 4G is that any organizations
and individuals can optionally arrange their AP (Wireless Access Point) according
to their own needs [66]. Due to the limitation of the spectrum and the randomness
of the AP arrangement, the interference in the network has become a major issue.
Take 802.11b/g IEEE series protocol as an example. The protocol defines
13 channels which can be used by a WLAN. Each channel occupies 22 MHz
bandwidth, but adjacent channel spacing is only 5 MHz. If the channel is completely
orthogonal, at least five channels must be spaced between adjacent channels [67].
Then the entire spectrum can support three APs’ orthogonal configurations at most.
In reality, a WLAN will firstly configure itself into an optimal network environment,
but in the use process, every AP channel may fluctuate. The research focus of the
paper lies in how to quickly detect the AP that has a problem in self-optimization.
110 4 Intelligent Network Control

4.3.1 Som Network Training

In order to enable the network to have some self-organization features to reduce

manual operations and maintenance [117], we use SOM to establish an artificial
neural network model [26, 31, 40, 73]. We make the SINR values of all points
which are simulated in space of software platform as the input vectors to study
and classify, list all cases triggering interference self-optimization, and extract
the vector characteristics which have trigger effects. a number of trainings are
carried out. Then, the feature vectors are trained. The SOM network can cluster all
characteristics of interference for self-optimization, which helps quickly determine
the location of the AP fault.
The steps of applying a SOM neural network to diagnose network faults are as
follows:
• Select reference fault samples.
• Learn each standard fault sample, and after learning, label the fault of the neuron
with the largest output value.
• Feed samples into the SOM neural network model.
• If the output neurons in the output layer are in the same position with a standard
fault sample, it indicates the sample has a corresponding fault; and if there are
output neurons which are in the input layer having the location between many of
the standard faults, indicating that these standards are all likely to occur, and the
degree of each fault is determined by the Euclidean distance between the location
of this fault and the corresponding standard fault location [74].

4.3.2 Som Neural Network Optimization

SOM Neural networks is proposed by a Finland scholar Ko-honen as a kind of

Self-organizing Artificial neural network. The successful application area of SOM
includes data mining, pattern recognition, signal processing and some other fields.
It is based on feature clustering and high dimensional visualization [42]. SOM is
generated by simulating competition, lateral inhibition which means neurons in each
layer of the network can inhibit other neurons responsive to the input mode in order
to make itself the winner, self-organization and other characteristics [30, 75].
Figure 4.11 shows the structure of SOM neural networks [76]. Each neuron of
the input layer collects input signals from each neuron of the output layer using
a weight vector. The output layer is also called a competitive layer. SOM neural
network can map the input signal matrix of an arbitrary dimension to the discrete
matrix of one-dimensional or two-dimensional, also can achieve this transformation
by the way of topological structure [119]. Each neuron of network layer competes
to corresponding opportunity, and finally, there are only one neuron in the victory
[44]. Also those connection weights related to the winning neuron are adjusted
toward more conducive to competition. Upon receiving a set of data, the SOM neural
networks can cluster the data automatically following a specific neuron.
4.3 WLAN Interference Self-Optimization Based SOM Neural Networks 111

Fig. 4.11 SOM neural network

SOM clustering algorithm learning process is as follows, all the formula is based
on SOM Neural Network reference [77]:
(a) Initializing network ζ0 , let n represent the dimension of the input data space,
and input vector is denoted by

X[x1 , x2 , x3 , . . . , xn ]T (4.30)

The number of synaptic weight vector equals to the dimension of input data.
The weight vector of neuron J is denoted by

vj = [vj 1 , vj 2 , vj 3 , . . . , vj n ] j ∈ζ (4.31)

(b) Similarity matching. Search n times for the best matching (winning) neuron β
by least Euclidean distance criteria.

β(X) = arg min |X − vj | j ∈ζ (4.32)

(c) Updating. Adjusting the connection between winning nodes β(X) and theirs
neighborhood nodes. Through reference [78], we can obtain the following
formula:


vj (n + 1) = vj (n) + η (n) hβ(x), j (n) x (n) − vj (n) (4.33)

−d 2
The neighborhood function hi(x),j is denoted by Gaussian function exp( 2σj,β
2 )
neighborhood radius σ (n) = σ exp( −nτ1 ) contains more adjacent nodes as far as
112 4 Intelligent Network Control

possible, τ1 is a time constant, σ0 is an initial neighborhood radius; dβ,j = |xj −xβ |2

is the lateral distance of winning nodes β and excitatory node j .
Adjusting learning parameter η(n) using Eq. (4.34).
 
−n
η(n) = η0 exp (4.34)
τ2

τ2 is another time constant of SOM clustering algorithm. Starting from the initial
value η(0), η(n) decreases gradually.
The algorithm details of the system model
First and foremost, the total model is based on the ICIC theory, connecting with
the SOM neural networks. There are mainly have two cardinal line,
• One is directing at the N Aps group, do optimization at the whole group. The aim
is to control the total power to be the most efficient generally.
• Another is directing to every separate AP, to control the own power of one AP.
The aim is to reduce the same-frequency interference and make the power to be
more efficient.
At the Beginning of the algorithm, we need to define different threshold Pi for
different base station. Different Pi are corresponding to the magnitude of power for
different situation of communication. For example, while considering a situation S1
with people distributed uniformly, the amount of access and SNR and other related
parameters determines a set of power thresholds Pi . Meanwhile, for a situation S2
with dense crowds, the parameters are changed, therefore the set of power thresholds
Pi is changed. Thus the basic feature is Pi is a set of value of power thresholds
generated by different situation. Before doing the neural network optimization, the
algorithm will ask the user to set different parameters α1 , α2 , α3 , . . . And determine
the value of Pi for different situation.
These situations may be controlled by the measurement of AP and the set
of parameters α1 , α2 , α3 , . . . But we will only discuss the situation of a specific
situation and take its set of Pi value as input then focus on the historical Pi value of
a single AP. We will also find different Pi as new Pi . The period of Pi update is a
longtime learning period.
The learning period T cannot be too long, since it will reduce the optimization
of the system a lot. Meanwhile the learning period cannot be too short either, which
will waste power consumption and storage. The learning period is an important
parameter. After each learning period, all the Pi will change, which means the
situation is changed. Thus, T determines the working efficiecy and the optimizing
efficiency [37, 46, 78].
We must emphasize that the learning period do not need to equal to switching
time. Since the changing frequency of Pi may far outweigh T, Pi may be switched
just-in-time (live-update) for a system that is flexible enough, once the situation is
changed and be sensed by AP, the system can switch Pi . However, this is situation
maybe rare in the whole SQM neural network, so it is not representative enough. In
simulation, we set the switching time to be one tenth of T.
4.3 WLAN Interference Self-Optimization Based SOM Neural Networks 113

For a live-update system, AP will optimize the power consumption correspond

to the threshold of Pi and achieve the minimum power consumption of single AP
while still need to guarantee the quality requirement of communication.
For the Situation of N Aps group:
After each single AP study period (T) is completed, the next step is to adjust
the N APs group power [48]. In this step, we must assurance the communication
quality. Then, considering the N APs group as a control unit and cooperate with
SDN theory [79], we use the SOM neural network algorithm to optimize two aspect
of one control unit:
• Undermine the co-channel interference.
• Optimize the group power.
The first one is to optimize co-channel interference. In many circumstances, the
scene outbreak near a AP will lead to the power surge. At the same time, surrounding
APs also be interfered stronger by the AP. Here, the scenario could let the APs
associated with the one have to raise the power [80]. However, different scenarios
results in different situations. A live-update system here using SOM neural networks
to optimize the co-channel interference, is the first important aspect of our software.
Secondly, it’s the whole group power of the N APs domain optimization.
Based on ICIC thoughts, we could consider sharing the complementary situation
in the frequency domain. As an example, considering the overall power of N
APs cluster region to achieve the overall optimal power. That is, in the situation
of communication assurance, to control the overall power and minimum values
[51, 81]. Throughout the model, we determined the frequency of center as well as
edge according to the idea of ICIC model. At the same time, we make the definition
of two significant factor: μ1 and μ2 . The center-frequency power factor is μ1 , and
the edge-frequency power factor is μ2 . In consideration of the overall center and
edge power distribution, we do the optimization at the N APs group. We want to get
the result is to achieve the optimal power distribution.
Figure 4.2 shows the SOM network training workflow more clearly. self-
configuration and power self-configuration.

Process of a learning period (T):

1. Input the historical data set.
2. Input different parameter for different scenarios.
3. Use SOM network to do feature selection.
4. Apply affinity propagation clustering algorithm to the data selected,
determined the Pi .
5. Update the data set, and check the communication situation.
6. Begin the N APs group co-channel interference optimizing.
7. According to the total communication situation, optimizing the total
power efficiency.
8. Check every AP communication situation.
9. Predict the output and do comparison.
114 4 Intelligent Network Control

Fig. 4.12 APs’ basic frequency distribution

Frequency Self-Configuration We take a more basic cellular structure to do the

frequency distribution of AP. We select 1, 6, 11 channel as the AP’s initial allocation
channel, distinguished by colors, which shows in Fig. 4.12.
Power Self-Configuration First, we calculate the reuse distance between the same
frequency, and then set the power to meet the requirements of the ratio of coverage,
and finally self-configuration is divided into two conditions—the highest Energy
Efficiency Ratio and the best Signal Interference Noise Ratio.
We enable the initial self-configuration of 19 APs, which is shown in Fig. 4.13:

4.3.3 Network Training And Simulation

4.3.3.1 Simulation Platform

Software Development Environment

• System: Linux (Ubuntu 14.04)
• CPU: 4 GHz
• Memory: 16 G
• Compiling environment: C++, qmake
• Packet: Qt-OpenGL
4.3 WLAN Interference Self-Optimization Based SOM Neural Networks 115

Fig. 4.13 The initial self-configuration of APs

Table 4.3 Specific functional architecture of software

Class Function
Ap The basic attributes of stored AP
Dot Some basic properties of stored map points
EventButton Events button, basic attributes of stored events
EventSet Read-in of parameters of event editor window
GraphicsTest Unit tests of the drawing component
Introduce Software introduction window
Main Start the interface
Mainwindow Main window
Self_optimize_set Read in the optimization of related parameters
Self_set Algorithm call interface
Self_youhua The realization of the self-optimization algorithm
Set_algorithm The implementation of the self-configuration algorithm
Timer Multithreading management of events
User Store the basic attributes of the user

Software is constructed using “qt” which is based on C++ graphical interface

application development framework, and the visual performance of AP uses the
qt—OpenGL library. The specific classes are listed in Table 4.3.
System simulated to implement the related functions of self-optimization, includ-
ing self-planning, self-configuration, sleep patterns, load balancing, and other issues
about cover and interference.
We validate the feasibility of the theory by software simulation platform,
evaluate the manifestation of our algorithm, and convert the result to the actual
116 4 Intelligent Network Control

communication network optimization scheme to provide operators with the base

station deployment, the related policy decisions.
The visualization implement of model uses the Qt—OpenGL drawing com-
ponent. In prophase, we thoroughly studied and compared the OpenGL, OpenNI
library and NiTE library, read the related English documents and got familiar
with the basic principle of qt graphical QML programming ideas of operation,
which helped to understand and call the relevant API. In view of the practical
problems, our software selectively uses part of the self-configuration as well as
the self-optimization to reach the purpose of judging the AP layout and allocating
appropriate power.

4.3.3.2 Frequency Simulation

First, we begin with the neural network initialization settings, namely the construc-
tion of input layer neurons. We configured the 19-AP in simulation according to the
ideal state of power which represents the optimum coverage without interference,
and then set the SINR value of the 400 * 400 small lattice which is used as
the initial reference vector of neural network learning. Feature vector training
set is shown in Fig. 4.14 with the attachment lines between points representing

Fig. 4.14 Feature vector training set

4.3 WLAN Interference Self-Optimization Based SOM Neural Networks 117

Fig. 4.15 Self-organized SOM

the resulting weight vectors. Furthermore, Fig. 4.15 is the SOM Neighborhood
Weighted Distance of different positions within the region [82]. Color difference
represents the discrepancy of neighborhood weighted distance value in exact area.
We plot a layer of SOM networks in Fig. 4.15, where the blue patches are
expressed as neurons, and the red lines represent the direct neighbor relationships
between them. We colored the neighbor blocks from black to yellow to show the
association between each neuron’s weight vector and its neighbors.
Then, we use the software simulation platform to randomly produce several
numbers of different sets of input planes as a series of different characteristic vector,
continuously input data to the network to do the mapping layer weights learning. We
selected eight random and different Euler Distance (SOM Neighborhood Weighted
Distance) corresponding to their input planes as shown in Fig. 4.16.
A set of subgraphs are generated. Each ith subgraph indicates the weights from
the ith input to the layer’s neurons. It displays black to represent zero connections,
and red to represent the strongest positive connections.
We consider the influence of the situation about any one of the 19 AP’s frequency
changing from 1 to 11 to all the surrounding areas (400 * 400 small lattice area).
Therefore, after 19 * 11 layers of iterative learning process, the SOM neural network
is built. The weighted position change of the characteristic value as shown in
Fig. 4.17.
In Fig. 4.17, green dots show the input vectors, and the blues are the weight
vectors of each neuron. This graph is represented by red lines that connect the
neighboring neurons to show how the SOM classifies the input space [83].
118 4 Intelligent Network Control

Fig. 4.16 SOM Neighbor weighted Distance according to different input planes

We further generate several multiple test sets in the simulation platform for the
neural network. We keep the initial conditions and use the platform to generate a
random band deployment, which is a simple hit to the wireless network. Here we
can change the frequency of the AP points from 1 to 11 in the band, for example,
as shown in Fig. 4.18, due to changes in the band, the software simulation platform
will automatically trigger interference from optimization.
Figure 4.19 shows abnormal clustering around the SINR. It can be observed that
the interference location occurs in the AP coordinates.
The results show that, in the event of co-channel interference, the SOM network
can quickly find the error situation, find the AP position of interference that occurs,
and determines the emergence of regional abnormal position (corresponding to the
position of a small grid). Then the network provides appropriate responses to the
user area [64, 84].
4.3 WLAN Interference Self-Optimization Based SOM Neural Networks 119

Fig. 4.17 Weighted position vector result after training

Hits
5

1 1
4

1
3

2
1

1
1

0 1 1 1

−1
−1 0 1 2 3 4 5 6

Fig. 4.18 Training result test

120 4 Intelligent Network Control

Fig. 4.19 Clustering results of SINR before and after exception

Power 1 (Max): Power 2: Power 3: Power 4:

a0<P1 <a1 a1<P1 <a2 a2<P1 <a3 a3<P1 <a4

Power 5: Power 6: Power 7: Power 8(Min):

a4<P1 <a5 a5<P1 <a6 a6<P1 <a7 a7<P1 <a8

Fig. 4.20 The degree of eight power intensity symbols

4.3.4 Power Efficiency Simulation

In Fig. 4.20, here, we define eight power ranges, defined by SOM cluster situations.
[85]
To power efficiency, the most important is to simulation a real situation. Here,
we randomly choose six APs as the maximum power at a time which is showed in
Fig. 4.21, then we do the total energy efficiency and let the network be more efficient
and available.
Here, as Step 1 is showed in Fig. 4.22, we can see that in the situation of the
maximum power, one AP cannot guarantee its cellar normal communication, there
has to be some neighborhood APs to help its communication. It can also be observed
that not every AP needs a neighborhood AP assistant, it can be considered by the
real threshold-the random Pi .
4.3 WLAN Interference Self-Optimization Based SOM Neural Networks 121

Exception Distribution

E E
4

E
3

2
E E

0 E

−1
−1 0 1 2 3 4 5 6

Fig. 4.21 Power efficiency of a random situation initialization

Exception Distribution

E E
4

E
3

2
E E

0 E

−1
−1 0 1 2 3 4 5 6

Fig. 4.22 Power efficiency of a random situation step 1

122 4 Intelligent Network Control

Exception Distribution

E E
4

E
3

2
E E

0 E

−1
−1 0 1 2 3 4 5 6

Fig. 4.23 Power efficiency of a random situation step 2

Step 2 showed in Fig. 4.23, the total network also provides the communication
guarantee, in the extreme random situation it needs to do two step communication
guarantee, we can see that some extreme situation also need to do second expansion
and use neighborhoods to support the total environment. Here, the total Pi still keeps
increasing.
Step 3 showed in Fig. 4.24, we can see that some of the extreme APs start
decreasing power, and all the network begins to optimize the power distribution.
In this step, all the APs have already considered the second neighborhood.
Step 4 shown in Fig. 4.25 and step 5 in Fig. 4.26 also do optimization to the whole
network, and step 5 is the last step.
From the comparison with the traditional way of power distribution shown
in Fig. 4.27, we can see that one AP extreme power cannot guarantee the cellar
communication. Also, one AP use extreme power can lead strong interference. The
different Pi , from here, we can define the standard in our software by ourselves
using SOM and historical data.
Next, we have a contrast with another extreme situation.
In random situation 2, 6 extreme situations gathered in the center of Fig. 4.28,
and we need to optimize the total AP. And we give the final result using SOM as
shown in Fig. 4.19.
By comparing the traditional results with the result using SOM way in Fig. 4.29,
we can see clearly that the SOM power efficiency is really available, both guarantee-
ing the communication quality, at the same time, considering the power efficiency
and the total network optimization.
4.3 WLAN Interference Self-Optimization Based SOM Neural Networks 123

Exception Distribution

E E
4

E
3

2
E E

0 E

−1
−1 0 1 2 3 4 5 6

Fig. 4.24 Power efficiency of a random situation step 3

Exception Distribution

E E
4

E
3

2
E E

0 E

−1
−1 0 1 2 3 4 5 6

Fig. 4.25 Power efficiency of a random situation step 4

124 4 Intelligent Network Control

Exception Distribution

E E
4

2
E E

0 E

−1
−1 0 1 2 3 4 5 6

Fig. 4.26 Power efficiency of a random situation step 5

Concentrate Exception Distribution

E E
3
E E

2
E E

−1
−1 0 1 2 3 4 5 6

Fig. 4.27 In traditional way of power distribution

4.3 WLAN Interference Self-Optimization Based SOM Neural Networks 125

Concentrate Exception Distribution

E E
3
E E

2
E E

−1
−1 0 1 2 3 4 5 6

Fig. 4.28 Random situation 2 initialization

Concentrate Exception Distribution

E E
3
E E

2
E E

−1
−1 0 1 2 3 4 5 6

Fig. 4.29 Random situation 2 result using SOM way

126 4 Intelligent Network Control

SOM Tradition

Original Exception Distribution 2 6

5 0 0
4 21
E E
4 1 0
12 7
E
8 2
3
3 0
5 0
2
E E

0 E

-1
-1 0 1 2 3 4 5 6

SOM Tradition

5 6
0 0
3 12
0 0
4 1
7 2
7 8
10 7
4.4 Blockchain-Based Software-Defined Industrial Internet of Things: A. . . 127

4.4 Blockchain-Based Software-Defined Industrial Internet

of Things: A Dueling Deep Q-Learning Approach

Recently, there are a growing number of applications that use Internet of Things
(IoT) technologies in several industries. Industrial Internet of Things (IIoT) has
emerged and attracted lots of attentions from industry and academia [91]. In order
to meet the demands of high bandwidth, ubiquitous accessibility, and dynamic
management, software-defined networking (SDN) [93] has been used in IIoT, called
SDIIoT [94]. In addition, software-defined routing management, edge computing,
flow scheduling, and energy harvesting have been researched in excellent literature
[95, 97–99]. With the increasing number of industrial devices, more than one
controller is employed in SDIIoT, known as distributed SDIIoT. How to reach
consensus among multiple controllers is challenging in distributed SDIIoT.
Although some traditional methods enable to reach consensus among multiple
controller instances, numerous non-trivial issues in the current consensus methods
prevent SDIIoT from being used as a generic platform for different services and
applications, including (1) the extra overheads, (2) the poor safety and liveness
properties, (3) the limited scope of available network size. These challenges need to
be broadly tackled by comprehensive research efforts.
Recently, Blockchain (BC) [35, 100] has been emerged as a novel technique,
which can be used to address the above challenges. BC is a distributed ledger to
record transactions, and provide trustworthy services to a group of nodes without
central authority [101]. For the distributed SDIIoT, BC can act as a trusted third and
‘out-of-band’ party to collect and synchronize network-wide views (e.g., network
events, network topology, and OpenFlow commands, etc) between different SDN
controllers safely, dependably, and traceably. In general, there are two kinds of BC
[103], including permissionless BC and permissioned BC. In permissionless BC,
enrollment is open to anyone, and nodes enable to join and leave dynamically and
frequently, using Nakamoto consensus protocol and coupled to cryptocurrency, such
as proof of work (PoW) in Bitcoin [100] and proof of stake (PoS) in Ethereum
[104]. All BC participants (miners) contribute their CPU power to work on an
extra hard task, and only the winner of them enables to propose a block and
synchronize it with others. Thus, lots of resources and time are imposed in the
permissionless BC. Moreover, another BC, the permissioned BC, uses Byzantine
fault tolerance (BFT) consensus protocol. It employs state machine replication
mechanism to deal with Byzantine nodes that are subverted by adversaries and
against the common goal of reaching consensus maliciously [105, 108], such as
practical Byzantine fault tolerance (PBFT) [107] and Paxos [3, 109]. They always
operate in a partially trusted environment, such as Hyperledger Fabric [110]. Thus,
the advantages of permissioned BC are low costs, low latency, and low band-
intensive. Considering partial trust, limited communication time, high loads and
narrow bandwidth in SDIIoT, as well the advantages of permissioned BC, we
consider to use permissioned BC in this part.
128 4 Intelligent Network Control

In this part, we propose a BC-based consensus protocol in distributed SDIIoT,

where BC works as a trusted third party to collect and synchronize network-wide
views between different SDN controllers. Specifically, it is a permissioned BC. Due
to the fact that the throughput of permissioned BC is constrained with respect to the
performance of BFT consensus protocol, we jointly consider the trust features of BC
nodes and controllers, as well as the computational capability of BC system, so as
to further improve the throughput of BC. Accordingly, we formulate view change,
access selection, and computational resources allocation as a joint optimization
problem. We describe this problem as a Markov decision process by defining state
space, action space, and reward function. Inspired by the works to achieve the
optimal decisions by improved learning methods [112–114], and the fact that it is
difficult to solve this joint problem by traditional methods, we use a novel dueling
deep Q-learning approach to solve this problem. A preliminary version of our work
appeared in [115]. The distinct contributions of this part are as follows.
• We propose a BC-based consensus protocol to simplify and secure the collection
and synchronization of network views between different SDN controllers. And
we give consensus steps in detail, along with theoretical analysis.
• Due to the fact that lots of network views need to be collected and synchronized
by BC system, we jointly consider the trust features of BC nodes and controllers,
as well as computational capability of the BC system, so as to improve the
throughput. Accordingly, we formulate view change, access selection, and
computational resources allocation as a joint optimization problem. We describe
it as a Markov decision process by defining state space, action space, and reward
function.
• In order to address this joint problem, we propose a novel dueling deep Q-
learning approach to learn the optimal strategy. Simulation results with different
system parameters are presented to show the effectiveness of the proposed
scheme.

4.4.1 System Model

In this subsection, we introduce the system model that we use. We first present the
network model, followed by the trust feature model and the computation model.

4.4.2 Network Model

We assume that there are C controllers in distributed SDIIoT, which are represented
by C = {1, . . . , C}. Each of them can communicate with the third-party BC
system. This BC system consists of N nodes, i.e., physical machines, denoted by
N = {1, . . . , N }. Like other robust BFT protocols [120], these N nodes are under
4.4 Blockchain-Based Software-Defined Industrial Internet of Things: A. . . 129

Fig. 4.30 The different network structures between traditional scheme and BC-based scheme

Byzantine failure model to make consensus, where at most f nodes are faulty [107],
where f = N 3−1 . And any finite number of controllers can behave arbitrarily to
issue correct or incorrect transactions to the BC system. Some strong adversaries can
collude with each other to compromise the replicated service. However, they can’t
break cryptographic technologies, i.e., signatures, message authentication code
(MAC), and collision-resistant hashing. We denote the messages with cryptographic
technologies as follows [38, 121].
• mσi means that message m is signed with a pubilc key from node i.
• mσi,j means that message m is authenticated by node i with a MAC for node j .
• mσi means that message m is authenticated by an array of MACs with node i
for every replicas.
Figure 4.30 shows the different network structures between traditional scheme
and BC-based scheme. It is worth mentioning that we use edge computing servers
to do some computations related to the above cryptography so as to improve the
throughput of BC system. There are E edge computing servers, and the set of
computing servers is represented by E = {1, 2, . . . , E}.

4.4.2.1 Trust Feature Model

We consider trust features of nodes and controllers in the system. Due to the
lack of centralized security services and prior security association, all nodes and
controllers have diverse trust features, such as safe or compromised. It is barely
possible to exactly know what the trust feature is for one node or one controller
at the next time instant. Thus, the trust features of a node n ∈ {1, 2, . . . , N } and
a controller c ∈ {1, 2, . . . , C} can be modelled as random variables δ n and ηc . δ n
and ηc can be divided into discrete levels, denoted by ξ = {ξ0 , ξ1 , . . . , ξL−1 }, and
130 4 Intelligent Network Control

D = {D0 , D1 , . . . , DH −1 }, respectively, where L and H are the number of available

trust features for nodes and controllers. We assume the trust features realization
of δ n and ηc to be δ n (t) and ηc (t) at time slot t, respectively. There are T time
slots during the period of time, which starts from when the controller issues an un-
validated block, and terminates when the controller is replied with a validated block.
Let t ∈ {0, 1, 2, . . . , T − 1} denote the time instant.
Considering the time correlation of real trust features in BC nodes and con-
trollers, we use Markov chain to model the transition of trust features in BC nodes
and controllers, as follows:
• For node n, let the transition probability of δ n (t) from one state Xs to another
state Ys be κXs Ys (t). The L × L transition probability matrix K n (t) of the trust
feature in node n is denoted as

K n (t) = [κXs Ys (t)]L×L , (4.35)

where κXs Ys (t) = P r(δ n (t + 1) = Ys |δ n (t) = Xs ), and Xs , Ys ∈ ξ .

• For controller c, the transition probability of ηc (t) is γθs φs (t). The H × H
transition probability matrix Υ c (t) of the trust feature in controller c is denoted as

Υ c (t) = [γθs φs (t)]H ×H , (4.36)

where γθs φs (t) = P r(ηc (t + 1) = φs |ηc (t) = θs ), and θs , φs ∈ D.

4.4.2.2 Computation Model

There are a number of computation tasks in the BC nodes, such as verifying

signatures, generating MACs, and verifying MACs. Let Tm = {sm , qm } denote a
computation task related to message m, where sm means the size of message m, and
qm is the required number of CPU cycles to complete this task.
In order to improve the throughput of BC, we use virtual computing resources
from edge computing servers to do computation tasks in the BC system. There are
lots of edge computing servers and some other computation tasks that also use the
computation resources in edge computing servers, thus we don’t exactly know the
computational resources for the BC system at the next time instant. Therefore, we
model the computation resources of edge computing server e for the BC system as a
random variable ζ e . To discretize the values of computation capabilities, ζ e can be
partitioned into Y discrete intervals as Y = {Y0 , Y1 , . . . , YY −1 }. The computation
resources of edge computing server e at time slot t can be denoted as ζ e (t), t ∈
{0, 1, 2, . . . , T − 1}. Considering the time correlation of computation state in edge
computing server, we use a Markov chain to model the transition of computation
state in edge computing server. Based on a certain transition probability, ζ e (t)
4.4 Blockchain-Based Software-Defined Industrial Internet of Things: A. . . 131

changes from one state to another. Let ϑas bs (t) denote the transition probability.
The Y × Y computation state transition probability matrix is represented as:

Π n (t) = [ϑas bs (t)]Y ×Y , (4.37)

where ϑas bs (t) = P r(ζ e (t + 1) = bs |ζ e (t) = as ), and as , bs ∈ Y .

The execution time of computation task Tm can be denoted as
qm
tm = . (4.38)
ζ e (t)

Thus, the computation rate is

sm ζ e (t)sm
CompR e (t) = a e (t) = a e (t) , (4.39)
tm qm

where a e (t) means whether or not edge computing server e is allocated to the BC
system at time slot t. a e (t) = 1 denotes edge computing server e is allocated to the
BC system; otherwise a e (t) = 0. At one time slot, there is only one edge computing
server allocated to the BC system, thus E e=1 a (t) = 1.
e

4.4.3 Blockchain-Based Consensus Protocol

We have presented that the existing consensus protocols are challenging in SDIIoT,
and BC could be a potential approach to address these issues in the previous
subsection. In this section, we propose a novel BC-based consensus protocol in
distributed SDIIoT. We begin with an overview of BC-based consensus protocol.
Then, we present the detailed steps of the consensus protocol, along with theoretical
analysis.

4.4.3.1 Overview of BC-Based Consensus Protocol

Each controller collects its local events and OpenFlow commands as Transaction
#1, Transaction #2, . . . , Transaction #n, which is called the collection period. The
format of a transaction is shown in Table 4.4. The number of this transaction denotes
the position of this transaction. The signature and MAC make sure the integrity and
authentication of this transaction. The payloads include local events and OpenFlow
commands that need to be synchronized between different SDN controllers.
After the collection period, all controllers issue consensus requests to the third-
party BC system. According to a policy, which will be introduced in the following
subsection, called access selection, the BC system only enables one controller to
access, and replies it by an admission message [39]. Then, this controller sends an
132 4 Intelligent Network Control

Table 4.4 The format of a The number of this transaction in the block
transaction
The Signature of this transaction
The MAC of this transaction
Payloads, including local events and OpenFlow commands

Table 4.5 The format of a block

Field Description
Version Block version number
Timestamp Creation time of this block
Controller ID The identifier of this controller
Block ID The identifier of this block
Block payload Transactions in this block (Transaction #1, . . . , Transaction #n)

Fig. 4.31 The overview of consensus procedures in BC-based consensus protocol between
different SDN controllers

un-validated block with block header and transactions, whose format is presented
in Table 4.5. After reaching consensus, the BC system sends the corresponding
validated block to the entire controllers. Finally, all controllers learn the payloads in
each transaction to know the events and OpenFlow commands from other controller.
These steps are in the consensus period. By this way, network-wide views can be
synchronized between different SDN controllers.
For a comprehensive perspective, Fig. 4.31 offers the consensus procedures in
blockchain structure. Here, controller 1 is the selected controller to access to the BC
system.
4.4 Blockchain-Based Software-Defined Industrial Internet of Things: A. . . 133

Fig. 4.32 The detailed procedures inside the permissioned BC

4.4.4 Detailed Steps and Theoretical Analysis

After giving the overview of BC-based consensus protocol, we will introduce the
detailed steps inside the permissioned BC, along with the theoretical analysis of
each step.
Based on PBFT [107], the detailed steps inside the permissioned BC is depicted
in Fig. 4.32. The numbering of each step in this figure is the same as the one used
in the remainder of this subsection. PBFT protocol has been used in real scenario,
such as Hyperledger Fabric project [41, 110], and Hyperledger Indy project [122].
They are hosted by The Linux Foundation, and develop applications with a modular
architecture. Thus, we consider that it can be used in real scenario when meeting
with SDN controllers.
1. The Controller Sends an Un-Validated Block to all Nodes The selected
controller sends an un-validated block to the BC system. The agent chooses one
node in the BC as the primary node p. Making decisions about which node is
the primary node is known as view change protocol. The view change protocol
used in our proposed scheme will be introduced in the following subsection. This
selected controller sends a block message blockσc , cσc to all nodes, where c
denotes the controller ID. It is encrypted with the private signature of controller
c, and authenticated with MACs for all nodes. When receiving this message, only
primary node p verifies the MAC. If valid, the signature will be verified then. If still
valid, it verifies the signature and the MAC of each transaction in this block, then
moves to the following steps. The success rate of all verification will be recorded
by the agent. If this block has already been executed, the primary will resend the
validated block to this controller.
134 4 Intelligent Network Control

Theoretical Analysis We consider an uncivil execution during which a fraction g

of transactions sent by the controller are correct [121]. The more trusted controller
has the bigger g. We assume verifying one signature, generating one MAC, and
verifying one MAC require θ , α, and α cycles, respectively, and the controller issues
batches of transactions of size b. As the work in [43, 107], we ignore the cost of
sending and receiving transactions. Therefore, the cost at the primary is

b
(1 + )(θ + α), (4.40)
g

and there is no cost at non-primary nodes.

2. The Primary Node Multicasts PRE-PREPARE Message to Other Replica
Nodes Finishing the verification, primary p sends a PRE-PREPARE message to all
other replica nodes, as P RE−P REP ARE, p, c, H (m)σp , which is authenticated
with MACs for each replica node. Here, p and H (m) mean the primary node ID
and the hashed result of the issued block, respectively. The replica node verifies the
MAC of primary p, as well as the signature and the MAC of each transaction. Then
it enters the following steps.
Theoretical Analysis In this phase, primary p generates (N − 1) MACs for all
replicas. Each replica verifies one MAC from primary p, as well as gb signatures
and MACs from transactions. Thus, the cost at the primary is

(N − 1)α, (4.41)

and the cost at each replica node is

b
α+ (θ + α). (4.42)
g

3. The Replicas Send PREPARE Message to Others After verifying the validity
of MACs and signatures, each replica replies the PRE-PREPARE message with
sending a PREPARE message to all nodes, as P REP ARE, p, c, H (m), nσn ,
where n denotes the replica node ID. When each replica node collects 2f matching
PREPARE messages with its local PRE-PREPARE message, it will enter the
following steps.
Theoretical Analysis In this phase, primary p needs to verify 2f MACs. Each
replica node generates (N − 1) MACs and verifies 2f MACs. Therefore, the cost at
the primary is

2f α, (4.43)

and the cost at each replica is

(N − 1 + 2f )α. (4.44)
4.4 Blockchain-Based Software-Defined Industrial Internet of Things: A. . . 135

4. All Nodes Send COMMIT Message to Others Following the reception of 2f

matching PREPARE messages, node n sends a COMMIT message to all others, as
COMMI T , p, c, H (m), nσn . After receiving 2f matching COMMIT messages,
it will enter the following steps.
Theoretical Analysis In this phase, primary p needs to generate (N − 1) MACs,
and verify 2f MACs. Each replica generates (N − 1) MACs, and verifies 2f MAC.
Therefore, the costs at the primary and the replica are both

(N − 1 + 2f )α. (4.45)

5. The Nodes Send the Validated Block to all Controllers Node n sends a REPLY
message REP LY, block, nσn,c to all controllers, where block is the validated
block. When each controller receives 2f valid and matching REPLY messages, it
accepts this validated block and updates the corresponding network views.
Theoretical Analysis In this phase, the primary and the replicas need to generate
b
g MACs for one controllers. Therefore, the total costs at the primary and the replica
are both
b
Cα. (4.46)
g

Thus, for one transaction, the cost at the primary is

1 1 1 C+1 2N + 4f − 2
( + )θ + ( + )α + α (4.47)
g b b g b

For one transaction, the cost at the replica is

1 2+C 2N + 4f − 2
θ+ α+ α (4.48)
g g b

We assume that multi-core computation modules run in parallel on distinct cores.

Each core has the computation speed of ϕ Hz. In addition, we consider an uncivil
execution where the primary is not fully trusted. The primary has k trust to affect the
system performance, and k ∈ [0, 1]. The more trusted primary node has the bigger
k. Therefore, the throughput of the consensus protocol is at most

kϕ
min[ 2N +4f −2
,
( g1 + 1
b )θ + ( b1 + C+1
g )α + b α
(4.49)
kϕ
2N +4f −2
]trx/s.
1
gθ + 2+C
g α + b α
136 4 Intelligent Network Control

4.4.5 Problem Formulation

We have presented the throughput of the BC system. In order to improve the

throughput, we need to make the joint decisions about view changes, access
selection, and computational resources allocation. In this subsection, we formulate
this joint problem as a Markov decision process by defining state space, action
space, and reward function.

4.4.5.1 State Space

In order to improve the throughput, the learning agent needs to sense state
s(t) at time slot t. As we have mentioned, the learning agent should make the
joint decisions about view changes, access selection, and computational resources
allocation. Accordingly, the learning agent needs to sense the trust features of
all nodes and controllers, as well as the computational capabilities of all edge
computing servers. Therefore, the state space can be represented as follows.
⎡ ⎤
δ 1 (t) δ 2 (t) . . . δ N (t)
s(t) = ⎣η1 (t) η2 (t) . . . ηC (t)⎦ . (4.50)
ζ 1 (t) ζ 2 (t) . . . ζ E (t)

4.4.5.2 Action Space

The agent mainly needs to decide view changes (i.e., which node is the primary
node), access selection (i.e., which controller can access to the BC system), and
computational resources allocation (i.e., which edge computing server should be
allocated to the BC system). Thus, the action space is denoted by

A(t) = {AN (t), AC (t), AE (t)}, (4.51)

where AN (t), AC (t), and AE (t) represent:

• AN (t) = [a 1 (t), a 2 (t), . . . , a n (t), . . . , a N (t)], which means whether or not node
n is the primary. And a n (t) ∈ {0, 1}, where a n (t) = 1 means node n is the
primary node, otherwiseit is the replica node. Note that the BC system only has
one primary node, thus N n=1 a (t) = 1.
n

• A (t) = [a (t), a (t), . . . , a c (t), . . . , a C (t)] decides which controller can

C 1 2

access to the BC system. And a c (t) ∈ {0, 1}, where a c (t) = 1 represents
controller c can access, otherwise a c (t) = 0. Note that  at onec time slot, only
one controller enables to access to the BC system, thus C c=1 a (t) = 1.
• AE (t) = [a 1 (t), a 2 (t), . . . , a e (t), . . . , a E (t)] determines which edge computing
server is allocated to the BC system. And a c (t) ∈ {0, 1}, where a e (t) = 1
denotes edge computing server e is allocated, otherwise a e (t) = 0. Similarly,
 E
e=1 a (t) = 1.
e
4.4 Blockchain-Based Software-Defined Industrial Internet of Things: A. . . 137

4.4.5.3 Reward Function

In order to improve the throughput, we model the throughput of the BC system as

reward function. According to (4.49), state space, and action space, we define the
reward function as:

k ϕ
r(t) = min[ 2N +4f −2
,
( g1 + b1 )θ + ( b1 + C+1
g )α + b α
(4.52)
kϕ
2N +4f −2
]trx/s,
1
g θ + 2+C
g α + b α

N E
where k = n n
n=1 a (t)δ (t), ϕ
= e e
e=1 a (t)CompR (t), and g
=
C c c
c=1 a (t)η (t).
Based on the above problem formulation, the learning agent senses state s(t)
at time slot t, and outputs a policy π that determines which action a(t) should be
taken. Then this action will be executed, i.e., one controller will enable to access to
the BC system, one node will be the primary node, and one edge computing server
will be allocated to the BC system. In order to let the learning agent remember the
experience and act better next time, the immediate reward r(t) will be fed back
to the learning agent. The trust features of nodes and controllers, as well as the
computational capabilities of edge computing servers change to next state s(t + 1).
Then the learning agent senses them and outputs another new policy, and so on so
forth [47]. The final goal is to achieve the maximum long-term reward. Summarily,
the interaction of the learning agent and the environment is shown in Fig. 4.33.

Fig. 4.33 The interaction of

the learning agent and the
environment
138 4 Intelligent Network Control

There are some challenges to solve the above problem formulation as follows.
1. The target is to maximize the long-term reward by step-and-step control.
However, the learning agent only senses the state at time slot t, and the action
taken at time slot t will affect the environment at time slot t + 1. The state
cannot be obtained ahead of time. Thus, the traditional optimization method,
only considering current state, is not feasible.
2. Considering the trust features of nodes and controllers, as well as the computa-
tional capabilities of edge computing servers, the system is high-dimensional and
high-dynamical. It is hard to make the joint and optimal decisions by traditional
methods.
3. In the BC system, taking which action has no relationship with what happens in
the next time slot. For example, when the learning agent selects one controller
to access to the BC system, its trust feature in the next time slot still changes
according to its transition probability matrix, not the action [49]. Therefore,
the traditional optimization method, learning the relationship between states and
actions, is not suitable.
To address the above challenges, we will propose a dueling deep Q-learning
approach in the next subsection to achieve the maximum long-term reward.

4.4.6 Dueling Deep Q-Learning

After the problem formulation, we consider a dueling deep Q-learning approach

to address this problem. In this subsection, we begin with the introduction of Q-
learning and deep Q-learning. Then we present dueling deep Q-learning that is used
in this section.

4.4.6.1 Q-Learning

In the Q-learning model, the agent interacts with the environment by perceptions
and actions. In one interaction step, the agent receives current state s(t) from the
environment, then selects an action a(t) as the output, and the value of this action
is measured by a scalar reward r(t). This action generates next state s(t + 1). The
agent selects actions to obtain the maximum long-term rewards. It learns to do this
over several interaction steps by systematic trials and errors, guided by Q-learning
[123]. Q-learning is a model-free algorithm using delay rewards. It aims to find a
policy π , mapping states and actions, to maximize the long-term rewards.
There are two popular approaches to denote the feedback from each step in terms
of long-term rewards, namely state-value function V π (s) and action-state value
function Qπ (s, a). V π (s) means the expected total reward in state s:
∞

V π (s) = E π [ γ k rt+k+1 |st = s], (4.53)
k=1
4.4 Blockchain-Based Software-Defined Industrial Internet of Things: A. . . 139

where E π [∗] means mathematical expectation, rt+k+1 means the immediate reward
at time slot t + k + 1, and γ ∈ (0, 1) is discount factor to balance immediate reward
and future reward. Moreover, Qπ (s, a) denotes the expected total rewards in state s
and action a:
∞

Qπ (s, a) = E π [ γ k rt+k+1 |st = s, at = a]. (4.54)
k=1

Q-learning evaluates Q(s, a) by a temporal difference method as follows.

Q(s, a) ← Q(s, a) + α(r + γ max Q(s , a ) − Q(s, a)), (4.55)

a

where α is learning rate, and α ∈ (0, 1]. The action with maximum Q(s, a) may be
chosen by the agent at each step. In the traditional Q-learning, each Q(s, a) is put
into Q-table. However, with the rapid increase of data dimension, it is challenging
to put all Q(s, a) into Q-table.

4.4.6.2 Deep Q-Learning

The rise of deep learning has provided a new tool to overcome the challenges. The
most important property of deep learning is that deep networks enable to find the
low-dimensional features of high-dimensional data by crafting weights and biases
in deep networks. Therefore, many researches have advocated to use deep networks
to approximate Q(s, a) instead of Q-table, i.e., Q(s, a, ω) ≈ Q(s, a), where ω is
the set of weights and biases in deep networks [124]. This is the core idea of deep
Q-learning (DQL).
In order to address the fundamental instability problem of approximating
Q(s, a), there are two improvements in DQL, including experience replay and
fixed target networks: (1) Experience replay stores the transitions as a set of
{state, action, reward, statenext } in a finite-sized cyclic buffer, and the agent
randomly samples batches of them to train deep networks, instead of only the
current ones. By this way, the temporal correlations that can adversely affect DQL
are broken. (2) Fixed target networks have the same architecture as the evaluated
ones, but are kept frozen for a period of time. The evaluated networks are trained
in each step to minimize loss function L(ω) to evaluate real Q(s, a), and L(ω) is
represented as

L(ω) = E[(r + γ maxa Q(s , a , ω− ) − Q(s, a, ω))2 ], (4.56)

where ω− is the weights and biases set in target networks, and ω is the weights and
biases set in evaluated networks. During training, the weights and biases in target
networks are updated with evaluated networks periodically. For a comprehensive
perspective, we present the workflows of DQL in Fig. 4.34.
140 4 Intelligent Network Control

Fig. 4.34 The workflows of DQL

Fig. 4.35 Dueling architecture

4.4.6.3 Dueling Deep Q-Learning Approach

However, for the majority of states in our system, the choice of actions in the agent
has no repercussion with what happens, i.e., actions have no relationship with states.
According to the work in [58, 125], dueling DQL is more efficient than natural
DQL. Based on this approach, some training processes have been carried out in real
scenario, such as drive games [125], which are simulated with better performance.
In the dueling DQL, there is another value function, A(s, a), which represents the
relative advantage of a action, called state-action value function. Learning A(s, a)
is easier to know which action has better consequences. Instead of one single stream
following the output layer of deep networks, there are two separate streams in
dueling DQL, where one computes state-value function V (s), and another computes
state-action value function A(s, a), called dueling architecture as shown in Fig. 4.35.
4.4 Blockchain-Based Software-Defined Industrial Internet of Things: A. . . 141

Finally, these two streams are aggregated as a output Q(s, a). This combination
module is denoted as follows:

Q(s, a; ω, ρ, ζ ) = V (s; ω, ρ) + A(s, a; ω, ζ ), (4.57)

where V (s; ω, ρ) is a scalar, and A(s, a; ω, ζ ) is an |A |−dimensional vector. ρ and

ζ are the parameters of two separate streams.

Algorithm 4 Dueling DQL

1: Initialization:
2: Initialize evaluated deep networks with weights and biases set ω.
3: Initialize target deep networks with weights and biases set ω− .
4: for k = 1 : K do
5: Reset the environment with a randomly initial observation sini , and s(t) = sini .
6: s(t)! = sterminal
7: Select action a(t) based on −greedy policy.
8: Obtain immediate reward r(t) and next observation s(t + 1).
9: Store experience (s(t), a(t), r(t), s(t + 1)) into experience replay memory.
10: Randomly sample some batches of (s(i), a(i), r(i), s(i + 1)) from experience replay memory.
11: Calculate two streams of evaluated deep networks, including V (s; ω, ρ) and A(s, a; ω, ζ ),
and combine them as Q(s, a; ω, ρ, ζ ) using (4.59).
12:
13: Calculate target Q-value Qtarget (s) in target deep networks:
14: if s is sterminal
15: Qtarget (s) = rs ,
16: else
17: Qtarget (s) = rs + γ maxa Q(s , a ; ω , ρ , ζ ).
18: Train evaluated deep networks to minimize loss function L(w)
19:

L(ω, ρ, ζ ) = E[(Qtarget (s) − Q(s, a; ω, ρ, ζ ))2 ]. (4.58)

20: Every some steps, update target deep networks.

21: s(t) ← s(t + 1)
22: end while
23: end for

According to the work in [125], considering the unidentifiability of (4.57), we

replace (4.57) as:

Q(s, a; ω, ρ, ζ ) = V (s; ω, ρ)+

1  (4.59)
(A(s, a; ω, ζ ) − A(s, a ; ω, ζ )).
|A |
a

As the output of dueling architecture is Q(s, a), it can be trained by many

existing algorithms. In this section, we use dueling architecture in natural DQL.
Therefore, the pseudocode of dueling DQL is presented in Algorithm 1, where
−greedy policy is used to balance the exploitation and the exploration.
142 4 Intelligent Network Control

4.4.7 Simulation Results and Discussions

In this subsection, we use computer simulation to evaluate the performance of our

proposed scheme. First, we present simulation settings, followed by simulation
results and the corresponding discussions.

4.4.7.1 Simulation Settings

In this simulation, hardware environment is a GPU-based server, and this server

has 8 GB 1867 MHz LPDDR3, 2 GHz Intel Core i5, and 256 G memory. Software
environment is Python 2.7.10 with TensorFlow 1.4.0 [126]. These two kinds of
simulation tools have been widely used from business to academic. TensorFlow
is a flexible, high-performance serving system, used by machine learning mod-
els, and designed for production environments. TensorFlow enables to keep the
same server architecture and application program interface (APIs), when different
machine learning algorithms are deployed. Thus, it has been widely used to deploy
new machine learning algorithms and experiments. We use these real simulation
environments, thus we can make sure that the performance of simulation results can
estimate and approximate the performance in real scenario.
We assume that there are four BC nodes, two controllers, and two edge
computing servers. The trust feature of each node is very safe, safe, medium,
compromised, and very compromised, whose transition probability matrix is
⎡ ⎤
0.5 0.15 0.125 0.12 0.105
⎢ 0.15 0.105⎥
⎢ 0.5 0.125 0.12 ⎥
⎢ ⎥
K = ⎢0.105 0.15 0.5 0.125 0.12 ⎥ . (4.60)
⎢ ⎥
⎣0.105 0.12 0.125 0.5 0.15 ⎦
0.105 0.12 0.125 0.15 0.05

Similarly, the trust feature of each controller can be very safe, safe, medium,
compromised, and very compromised. We set the transition probability matrix as
⎡ ⎤
0.45 0.16 0.14 0.13 0.12
⎢0.16 0.12⎥
⎢ 0.45 0.14 0.13 ⎥
⎢ ⎥
Υ = ⎢0.12 0.16 0.45 0.14 0.13⎥ . (4.61)
⎢ ⎥
⎣0.12 0.13 0.16 0.45 0.14⎦
0.12 0.13 0.14 0.16 0.45

We assume the computational capability of each edge computing server is high,

medium, low, and very low, and set the transition probability as
4.4 Blockchain-Based Software-Defined Industrial Internet of Things: A. . . 143

⎡ ⎤
0.5 0.3 0.15 0.05
⎢ 0.3 0.5 0.15 0.05⎥
Π =⎢
⎣0.15
⎥. (4.62)
0.3 0.5 0.05⎦
0.15 0.3 0.5 0.05

The values of the rest of parameters are summarized in Table 4.6. We use
TensorBoard to visualize the TensorFlow graph, as shown in Fig. 4.36.
For the performance comparison, there are four schemes simulated:
• Proposed dueling DQL-based scheme with view changes, access selection, and
edge computing servers. We call it duelingDQL-based scheme in the remainder
of this section. In other words, in duelingDQL-based scheme, the learning agent
enables to select more trusted BC node as the primary, more trusted controller
to access to the BC system, and edge computing server with more computing
capabilities. Thus, this scheme should have the best performance.

Table 4.6 Parameters setting in the simulation

Parameter Value Description
θ 8 Mcycles The required number of CPU cycles to verify one signature
α 0.05 Mcycles The required number of CPU cycles to verify and generate
one MAC
b 1 Mb The batch size of a block
γ 0.9 The discount factor

Fig. 4.36 The visualized TensorFlow graph in TensorBoard

144 4 Intelligent Network Control

• Proposed dueling DQL-based scheme with view changes, edge computing

servers, but without access selection. We call it dueling DQL-based scheme
without controller choice in the remainder of this section. That is to say, in
this scheme, the learning agent only enables to select more trusted BC node as
the primary, and edge computing server with more computing capabilities, but
controllers access to the BC system randomly. Thus, some malicious controllers
may issue incorrect transactions to slow down the performance. Compared with
the duelingDQL-based scheme, the advantage of selecting trusted controller can
be indicated.
• Proposed dueling DQL-based scheme with access selection, edge computing
servers, but with the traditional view change protocol in [107]. We call it
duelingDQL-based scheme without node choice in the remainder of this sec-
tion. In other words, in this scheme, the learning agent only selects more
trusted controller to access to the BC system, and edge computing server with
more computing capabilities, but using the traditional view change protocol
without considering historical trust reputation. Thus, some malicious nodes
may be selected as the primary to weaken the throughput. Compared with the
duelingDQL-based scheme, the merit of selecting trusted BC node can be shown.
• Proposed dueling DQL-based scheme with view changes, access selection,
but without edge computing servers, which only uses the local computing
capabilities in the BC system. We call it duelingDQL-based scheme without
computation offloading in the remainder of this section. That is to say, the
learning agent only selects more trusted BC node as the primary, and more
trusted controller to access to the BC system, but only using the local computing
capabilities. Without the assistance of edge computing servers, it is difficult to
do cryptographic tasks only by local computation, which may slow down the
performance. Compared with the duelingDQL-based scheme, the advantage of
using edge computing servers can be indicated.
• Existing scheme with traditional view changes, without access selection, and
with local computational capabilities. We call it existing scheme in the remainder
of this section. The dueling DQL approach is not employed in this scheme.
This scheme allows controllers to access to the BC system randomly, uses the
traditional view change protocol and local computing capabilities. Compared
with the duelingDQL-based scheme, the advantage of using the dueling DQL
approach can be shown.

4.4.7.2 Simulation Results

Figure 4.37 shows the relationship between training episodes and the throughput of
the BC system under different schemes. Each point is the average throughput per
episode. The agent runs in AdamOptimizer [127] with the learning rate of 1e−5 .
As we can see from this figure, with the joint consideration of node’s trust feature,
controller’s trust feature, and offloading the computation task to edge computing
servers, the BC system has the better performance. The reason is that the more
4.4 Blockchain-Based Software-Defined Industrial Internet of Things: A. . . 145

Fig. 4.37 Training curves tracking the throughput of the BC system under different schemes

trusted node is less possible to slow the system performance down, the more trusted
controller issues the higher fraction of correct transactions, and with the help of edge
computing servers, computation tasks can be executed more quickly. This figure
shows the convergence performance of dueling DQL as well. At the beginning of
learning and training, dueling DQL takes some trials and errors. With the increase
of episodes, the throughput turns to be stable, which means the agent has learned
the optimal policies to maximize the long-term rewards.
In addition, Fig. 4.38 shows the relationship between the learning loss in (4.56)
and the training steps of DuelingDQL-based scheme, when the agent runs in the
same parameters as above. At the beginning of learning, deep networks have no
knowledge of the uncertain environment, and with the increase of new experiences,
the learning loss is higher and higher. When the cyclic buffer of experiences in
dueling DQL is full, the agent has some knowledge of the environment, which leads
to the decrease of the learning loss. Such increasing and decreasing of the learning
loss indicate the effectiveness of deep networks.
Figure 4.39 shows the relationship between training episodes and the throughput
under different learning rates in Dueling DQL-based scheme. As we can see from
this figure, the learning rate has effects on the convergence performance. The
learning rate means the length of learning step to minimize the loss function. The
bigger learning rate denotes the longer learning step. The longer learning steps
are likely to miss the global optimum, which leads to the highly scaled curves
when learning rates are 0.01 and 0.001. The shorter learning steps may lead to
the slower convergence speed, because more steps are necessary to achieve the
146 4 Intelligent Network Control

Fig. 4.38 Training curves tracking the learning loss under DuelingDQL-based scheme

x104
4

3.5 DuelingDQL-based scheme

DuelingDQL-based scheme without controller choice
DuelingDQL-based scheme without node choice
3 DuelingDQL-based scheme without computation offloading
Existing scheme
Throughput (trx/s)

2.5

1.5

0.5

0
10 12 14 16 18 20 22 24 26 28 30
The number of controllers

Fig. 4.39 Training curves tracking the throughput of the BC system under different learning rates
4.4 Blockchain-Based Software-Defined Industrial Internet of Things: A. . . 147

x104
4.5

3.5 DuelingDQL-based scheme

DuelingDQL-based scheme without controller choice
DuelingDQL-based scheme without node choice
3
Throughput (trx/s)

DuelingDQL-based scheme without computation offloading

Existing scheme
2.5

1.5

0.5

0
6 8 10 12 14 16 18 20
The number of consensus nodes

Fig. 4.40 Training curves tracking the learning loss of natural DQL and dueling DQL

global optimum. Compared with two curves of blue and orange, although the orange
one has the faster convergence speed, its curve is unstable after the convergence.
Therefore, we choose the learning rate as 1e−5 in the simulation. Because its
convergence speed is acceptable and it has better learning stability.
Figure 4.40 shows the learning loss of natural DQL and dueling DQL. As we
can see, the learning loss in dueling DQL decreases more quickly than natural
DQL, which indicates dueling DQL has better learning effectiveness. The reason
is that in our BC system, the choices of which node is the primary, which controller
can access to the BC system, and which edge computing server should execute
the computation tasks have no relationship with states. Learning which action has
better consequence is more efficient than learning which state is better. In dueling
DQL, one stream learns state-action value function A(s, a), which is more useful to
help the agent make the good choices. Therefore, the learning loss in dueling DQL
decreases more fast than natural DQL.
After the effective training of deep networks, we use them in the following
simulations. Figure 4.41 shows the relationship between the number of controllers
and the system throughput under different schemes. This figure also can be used to
show the performance of these schemes in large real SDIIoT environment, where
up to 30 controllers are considered in this simulation. As the increase of the number
of controllers, the throughput of the BC system decreases. The reason is that more
controllers need more computational operations about verifying signatures, MACs.
But with the joint consideration of trust features of controllers and nodes, as well as
148 4 Intelligent Network Control

x104
5

4.5

4 DuelingDQL-based scheme
DuelingDQL-based scheme without controller choice
3.5 DuelingDQL-based scheme without node choice
Throughput (trx/s)

DuelingDQL-based scheme without computation offloading

Existing scheme
3

2.5

1.5

0.5
1 2 3 4 5 6 7 8 9 10
The batch size of a block

Fig. 4.41 The throughput versus the number of controllers under different schemes

using edge computing servers, our proposed scheme, as shown in the blue curve,
has better performance. Thus, we can see our proposed scheme still has better
performance in large SDIIoT environment.
Figure 4.42 shows the relationship between the number of BC nodes and the
system throughput under different schemes. As we can see, more nodes lead to the
less system throughput. The reason is that with the increase of the number of nodes,
more signatures and MACs need to be verified and generated, which need more
CPU cycles so as to decrease the system throughput. But, the performance of our
proposed scheme is still the best.
Figure 4.43 shows the relationship between the batch size of a block and the
system throughput. The bigger block enables to contain more transactions so as
to synchronize more local network events among controllers, which increases the
system throughput. As we can see, our scheme also has the better performance.

4.5 Summary

In this chapter, we discuss the main challenge of network control and introduced
several machine learning based algorithms. We first present an energy-aware multi-
controller placement scheme as well as a latency-aware resource management
model for the SDWN. Then, we present a novel controller mind (CM) framework to
4.5 Summary 149

x104
4.5

3.5 DuelingDQL-based scheme

DuelingDQL-based scheme without controller choice
DuelingDQL-based scheme without node choice
3
Throughput (trx/s)

DuelingDQL-based scheme without computation offloading

Existing scheme
2.5

1.5

0.5

0
6 8 10 12 14 16 18 20
The number of consensus nodes

Fig. 4.42 The throughput versus the number of BC nodes under different schemes

x104
5

4.5

4 DuelingDQL-based scheme
DuelingDQL-based scheme without controller choice
3.5 DuelingDQL-based scheme without node choice
Throughput (trx/s)

DuelingDQL-based scheme without computation offloading

Existing scheme
3

2.5

1.5

0.5
1 2 3 4 5 6 7 8 9 10
The batch size of a block

Fig. 4.43 The throughput versus the batch size of a blocks under different schemes
150 4 Intelligent Network Control

implement automatic management among multiple controllers and propose a novel

Quality of Service (QoS) enabled load scheduling algorithm based on reinforcement
learning to solve the problem of complexity and pre-strategy in the networks.
In addition, we present a Wireless Local Area Networks (WLAN) interference
self-optimization method based on a Self-Organizing Feature Map (SOM) neural
network model. Finally, we use a novel dueling deep Q-learning approach to solve
this joint problem in distributed SDIIoT.

References

1. S. Tamoor-Ul-Hassan, S. Samarakoon, M. Bennis, M. Latva-Aho, and C. S. Hong, “,” IEEE

Communications Letters, vol. 22, no. 1, pp. 137–140, 2018.
2. M. Huang, B. Yu, and S. Li, “Puf-assisted group key distribution scheme for software-defined
wireless sensor networks,” IEEE Communications Letters, vol. 22, no. 2, pp. 404–407, 2018.
3. M. Yue, C. Jiang, H. H. Chen, and R. Yong, “Cooperative device-to-device communications:
Social networking perspectives,” IEEE Network, vol. 31, no. 3, pp. 38–44, 2018.
4. H. Yao, C. Fang, Y. Guo, C. Zhao, H. Yao, C. Fang, Y. Guo, and C. Zhao, “An optimal routing
algorithm in service customized 5g networks,” Mobile Information Systems, vol. 2016, pp.
1–7, 2016.
5. B. Heller, R. Sherwood, and N. McKeown, “The controller placement problem,” in The 1st
Workshop on Hot Topics in Software Defined Networks, Helsinki, Finland, Aug. 2012, pp.
7–12.
6. M. J. Abdel-Rahman, E. A. Mazied, A. MacKenzie, S. Midkiff, M. R. Rizk, and M. El-Nainay,
“On stochastic controller placement in software-defined wireless networks,” in IEEE Wireless
Communications and Networking Conference (WCNC), San Francisco, CA, 2017, pp. 1–6.
7. H. Yao, Y. Hao, A. Zhang, F. Chao, and Y. Guo, “Wlan interference self-optimization using
som neural networks,” Concurrency & Computation Practice & Experience, vol. 29, 2017.
8. R. Grandl, G. Ananthanarayanan, S. Kandula, S. Rao, and A. Akella, “Multi-resource packing
for cluster schedulers,” ACM SIGCOMM Computer Communication Review, vol. 44, no. 4,
pp. 455–466, 2015.
9. C. Fang, H. Yao, Z. Wang, W. Wu, X. Jin, and F. R. Yu, “A survey of mobile information-
centric networking: Research issues and challenges,” IEEE Communications Surveys &
Tutorials, 2018.
10. V. Jalaparti, P. Bodik, I. Menache, S. Rao, K. Makarychev, and M. Caesar, “Network-
aware scheduling for data-parallel jobs: Plan when you can,” ACM SIGCOMM Computer
Communication Review, vol. 45, no. 4, pp. 407–420, 2015.
11. J. Du, C. Jiang, W. Jian, Y. Shui, H. Zhu, and R. Yong, “Resource allocation in space multi-
access systems,” IEEE Transactions on Aerospace & Electronic Systems, vol. 53, no. 2, pp.
598–618, 2017.
12. H. Mao, M. Alizadeh, I. Menache, and S. Kandula, “Resource management with deep
reinforcement learning,” in The 15th ACM Workshop on Hot Topics in Networks, Atlanta,
USA, Nov. 2016, pp. 50–56.
13. H. Yao, Q. Chao, F. Chao, C. Xu, and F. R. Yu, “A novel framework of data-driven
networking,” IEEE Access, vol. 4, pp. 9066–9072, 2016.
14. S. Knight, H. X. Nguyen, N. Falkner, R. Bowden, and M. Roughan, “The internet topology
zoo,” IEEE Journal on Selected Areas in Communications, vol. 29, no. 9, pp. 1765–1775,
Spe. 2011.
15. C. Jiang, L. Kuang, H. Zhu, R. Yong, and L. Hanzo, “Information credibility modeling in
cooperative networks: Equilibrium and mechanism design,” IEEE Journal on Selected Areas
in Communications, vol. 35, no. 2, pp. 432–448, 2017.
References 151

16. J. Gao, Y. Xiao, J. Liu, W. Liang, and C. P. Chen, “A survey of communication/networking in

smart grids,” Future Generation Comp. Sys., vol. 28, no. 2, pp. 391–404, 2012.
17. P. Si, H. Yu, H. Yao, R. Yang, and Y. Zhang, “Dave: Offloading delay-tolerant data traffic to
connected vehicle networks,” IEEE Transactions on Vehicular Technology, vol. 65, no. 6, pp.
3941–3953, 2016.
18. N. McKeown, T. Anderson, H. Balakrishnan, G. Parulkar, L. Peterson, J. Rexford, S. Shenker,
and J. Turner, “Openflow: enabling innovation in campus networks,” ACM SIGCOMM Comp.
Comm. Review, vol. 38, no. 2, pp. 69–74, 2008.
19. L. Xu, C. Jiang, Y. Qian, Y. Zhao, J. Li, and Y. Ren, “Dynamic privacy pricing: A multi-armed
bandit approach with time-variant rewards,” Trans. Info. For. Sec., vol. 12, no. 2, pp. 271–285,
Feb. 2017.
20. Y. He, F. R. Yu, N. Zhao, H. Yin, and R. C. Qiu, “Big data analytics in mobile cellular
networks,” IEEE Access, vol. 4, pp. 1985–1996, 2017.
21. S. Al-Rubaye, E. Kadhum, Q. Ni, and A. Anpalagan, “Industrial internet of things driven by
SDN platform for smart grid resiliency,” IEEE Internet of Things Journal, 2017.
22. M. Karakus and A. Durresi, “A survey: Control plane scalability issues and approaches in
software-defined networking (SDN),” Comp. Net., vol. 112, pp. 279–293, 2017.
23. C. Jiang, H. Zhang, H. Zhu, R. Yong, V. C. M. Leung, and L. Hanzo, “Information-
sharing outage-probability analysis of vehicular networks,” IEEE Transactions on Vehicular
Technology, vol. 65, no. 12, pp. 9479–9492, 2016.
24. H. Yao, Q. Wang, L. Wang, P. Zhang, M. Li, and Y. Liu, “An intrusion detection framework
based on hybrid multi-level data mining,” International Journal of Parallel Programming, pp.
1–19, 2017.
25. C. Qiu, C. Zhao, F. Xu, and T. Yang, “Sleeping mode of multi-controller in green software-
defined networking,” EURASIP Journal on Wireless Comm. and Net., vol. 2016, no. 1, pp.
282–296, Jan. 2016.
26. T. Zhao, H. Yao, C. Fang, J. Zhang, and P. Zhang, “Microblogging sentiment analysis method
with the combination of semantic rules and emoticon weighting,” Journal of Chongqing
University of Posts & Telecommunications, 2016.
27. E. Ng, Z. Cai, and A. Cox, “Maestro: A system for scalable openflow control,” Rice
University, Houston, TX, USA, TSEN Maestro-Techn. Rep, TR10-08, 2010.
28. S. Schmid and J. Suomela, “Exploiting locality in distributed sdn control,” in Proc. Conf. ACM
SIGCOMM Workshop on Hot Topics in Software Defined Net., Hong Kong, China, Aug. 2013,
pp. 121–126.
29. J. Wang, C. Jiang, H. Zhu, R. Yong, and L. Hanzo, “Network association strategies for an
energy harvesting aided super-wifi network relying on measured solar activity,” IEEE Journal
on Selected Areas in Communications, vol. 34, no. 12, pp. 3785–3797, 2016.
30. C. Liang, F. Yu, H. Yao, and H. Zhu, “Virtual resource allocation in information-centric
wireless virtual networks,” IEEE Transactions on Vehicular Technology, vol. 65, no. 12, pp.
1–1, 2016.
31. H. Yao, B. Zhang, P. Zhang, and M. Li, “A novel kernel for text classification based on
semantic and statistical information,” Computing and Informatics, vol. 37, no. 4, pp. 992–
1010, 2018.
32. M. Canini, P. Kuznetsov, D. Levin, and S. Schmid, “A distributed and robust sdn control plane
for transactional network updates,” in Proc. Conf. IEEE INFOCOM’15, Hong Kong, China,
Apr. 2015, pp. 190–198.
33. S. Civanlar, M. Parlakisik, A. M. Tekalp, B. Gorkemli, B. Kaytaz, and E. Onem, “A qos-
enabled openflow environment for scalable video streaming,” in Pro. Conf. GLOBECOM’10
Workshops, Miami, USA, Dec. 2010, pp. 351–356.
34. N. T. Hai and D.-S. Kim, “Efficient load balancing for multi-controller in SDN-based mission-
critical networks,” in Pro. Conf. Industrial Informatics’16, Poitiers, France, Jul, 2016, pp.
420–425.
35. P. Zhang, H. Yao, F. Chao, and Y. Liu, “Multi-objective enhanced particle swarm optimization
in virtual network embedding,” Eurasip Journal on Wireless Communications & Networking,
vol. 2016, no. 1, p. 167, 2016.
152 4 Intelligent Network Control

36. M. Yue, C. Jiang, X. Lei, R. Yong, and H. Zhu, “User association in heterogeneous networks:
A social interaction approach,” IEEE Transactions on Vehicular Technology, vol. 65, no. 12,
pp. 9982–9993, 2016.
37. P. Zhang, H. Yao, and Y. Liu, “Virtual network embedding based on computing, network, and
storage resource constraints,” IEEE Internet of Things Journal, vol. 5, no. 5, pp. 3298–3304,
2018.
38. F. Xu, H. Yao, C. Zhao, and Q. Chao, “Towards next generation software-defined radio access
network–architecture, deployment, and use case,” Eurasip Journal on Wireless Communica-
tions & Networking, vol. 2016, no. 1, p. 264, 2016.
39. L. Meng, F. R. Yu, P. Si, E. Sun, and H. Yao, “Random access and virtual resource allocation in
software-defined cellular networks with machine-to-machine (m2m) communications,” IEEE
Transactions on Vehicular Technology, vol. 66, no. 7, pp. 6399–6414, 2017.
40. J. Wang, C. Jiang, T. Q. S. Quek, X. Wang, and R. Yong, “The value strength aided
information diffusion in socially-aware mobile networks,” IEEE Access, vol. 4, pp. 3907–
3919, 2016.
41. H. Yao, F. Chao, Q. Chao, C. Zhao, and Y. Liu, “A novel energy efficiency algorithm in green
mobile networks with cache,” Eurasip Journal on Wireless Communications & Networking,
vol. 2015, no. 1, pp. 1–9, 2015.
42. J. Wang, C. Jiang, B. Zhi, T. Q. S. Quek, and R. Yong, “Mobile data transactions in device-
to-device communication networks: Pricing and auction,” IEEE Wireless Communications
Letters, vol. 5, no. 3, pp. 300–303, 2017.
43. H. Yao, C. Qiu, C. Zhao, and L. Shi, “A multicontroller load balancing approach in software-
defined wireless networks,” International Journal of Distributed Sensor Networks, vol. 2015,
no. 2, p. 10, 2015.
44. J. Wang, C. Jiang, H. Zhang, R. Yong, and V. C. M. Leung, “Aggressive congestion control
mechanism for space systems,” IEEE Aerospace & Electronic Systems Magazine, vol. 31,
no. 3, pp. 28–33, 2017.
45. H. Yao, C. Qiu, C. Zhao, and L. Shi, “A multicontroller load balancing approach in software-
defined wireless networks,” International Journal of Distributed Sensor Networks, vol. 11,
no. 10, pp. 41–49, 2015.
46. X. Lei, C. Jiang, R. Yong, and H. H. Chen, “Microblog dimensionality reduction—a deep
learning approach,” IEEE Transactions on Knowledge & Data Engineering, vol. 28, no. 7,
pp. 1779–1789, 2016.
47. However, “Modeling energy-delay tradeoffs in single base station with cache,” International
Journal of Distributed Sensor Networks, vol. 2015, pp. 1–5, 2015.
48. Y. Shen, C. Jiang, T. Q. S. Quek, and R. Yong, “Location-aware green communication design:
Exploration and exploitation on energy,” IEEE Wireless Communications, vol. 23, no. 2, pp.
46–52, 2016.
49. H. Yao, H. Tao, C. Zhao, X. Kang, and Z. Liu, “Optimal power allocation in cognitive
radio based machine-to-machine network,” Eurasip Journal on Wireless Communications &
Networking, vol. 2014, no. 1, p. 82, 2014.
50. J. Qi and D. Wu, “Green energy management of the energy internet based on service
composition quality,” IEEE Access, 2018.
51. J. Du, C. Jiang, Q. Yi, H. Zhu, and R. Yong, “Resource allocation with video traffic prediction
in cloud-based space systems,” IEEE Transactions on Multimedia, vol. 18, no. 5, pp. 820–830,
2016.
52. Y. Zhang, R. Yu, M. Nekovee, Y. Liu, S. Xie, and S. Gjessing, “Cognitive machine-to-machine
communications: visions and potentials for the smart grid,” IEEE Net., vol. 26, no. 3, 2012.
53. S. Maharjan, Q. Zhu, Y. Zhang, S. Gjessing, and T. Basar, “Dependable demand response
management in the smart grid: A stackelberg game approach,” IEEE Trans. on Smart Grid,
vol. 4, no. 1, pp. 120–132, 2013.
54. W. Xia, Y. Wen, C. H. Foh, D. Niyato, and H. Xie, “A survey on software-defined networking,”
IEEE Comm. Surveys & Tutorials, vol. 17, no. 1, pp. 27–51, 2015.
References 153

55. X. Lei, C. Jiang, Y. Shen, T. Q. S. Quek, H. Zhu, and R. Yong, “Energy efficient d2d
communications: a perspective of mechanism design,” IEEE Transactions on Wireless
Communications, vol. 15, no. 11, pp. 7272–7285, 2016.
56. R. Chaudhary, G. S. Aujla, S. Garg, N. Kumar, and J. J. Rodrigues, “SDN-enabled multi-
attribute-based secure communication for smart grid in IIoT environment,” IEEE Trans. on
Industrial Infor., vol. 14, no. 6, pp. 2629–2640, 2018.
57. A. Montazerolghaem, M. H. Yaghmaee, and A. Leon-Garcia, “OpenAMI: Software-defined
AMI load balancing,” IEEE Internet of Things Journal, vol. 5, no. 1, pp. 206–218, 2018.
58. Y. J. Liu, T. Huang, J. Zhang, J. Liu, H. P. Yao, and R. C. Xie, “Service customized
networking,” Journal on Communications, 2014.
59. Y. Zhang, R. Yu, S. Xie, W. Yao, Y. Xiao, and M. Guizani, “Home M2M networks:
architectures, standards, and QoS improvement,” IEEE Comm Mag., vol. 49, no. 4, 2011.
60. C. Li, S. Tsao, M. C. Chen, Y. Sun, and Y. Huang, “Proportional delay differentiation service
based on weighted fair queuing,” in Pro. Conf. Comp. Comm. and Net., Las Vegas, USA, Oct.
2000, pp. 418–423.
61. SDNCTC, http://www.sdnctc.com/.
62. H. Zhang, C. Jiang, R. Q. Hu, and Q. Yi, “Self-organization in disaster resilient heterogeneous
small cell networks,” IEEE Network, vol. 30, no. 2, pp. 116–121, 2016.
63. http://www.sdnctc.com/download/resource_download/id/6.
64. H. Yao, C. Zhao, and Z. Zhou, “Location based spectrum sensing evaluation in cognitive radio
networks,” Eurasip Journal on Wireless Communications & Networking, vol. 2011, no. 1, pp.
1–7, 2011.
65. Q. Zhang, A. Riska, W. Sun, E. Smirni, and G. Ciardo, “Workload-aware load balancing for
clustered web servers,” IEEE Trans. on Parallel and Distributed Sys., vol. 16, no. 3, pp. 219–
233, 2005.
66. C System, Inc., “Cisco Visual Networking Index: Global Mobile Data Traffic Forecast
Update, 2015–2020 [J]”, White Paper Cisco Systems Inc. [EB/OL], Feb. 2016.
67. Teuvo Kohonen and Timo Honkela, “Kohonen Network” [J], Scholarpedia, pp. 83–100, Jan.
2007. doi:10.4249/scholarpedia.1568
68. RU Qiang and RONG Meng-tian, “Research on IEEE 802.11b WLAN Adjacent Channel
Interference”, Information Technology, vol. 12, pp. 15–17, 2017.
69. Kim K H and Kang G S, “Self-Reconfigurable Wireless Mesh Networks”, IEEE/ACM
Transactions on Networking (TON), vol. 19, no. 2, pp. 393–404, Apr. 2011.
doi:10.1109/TNET.2010.2096431
70. Silva M W R D and De Rezende J F, “TDCS: A New Mechanism for Automatic Channel
Assignment for Independent IEEE 802.11 Networks”, Ad Hoc Networking Workshop, 8th
IFIP Annual Mediterranean, pp. 27–33, 2009.
71. Leong Yeng Weng, Jamaludin Bin Omar, Yap Keem Siah, Izham Bin Zainal Abidin, and Syed
Khaleel Ahmed, “Improvement of ANN-BP by Data Pre-Segregation Using SOM”, IEEE
International Conference on Computational Intelligence for Measurement Systems and Appli-
cation (CIMSA 2009), pp.175–178, May 11–13, 2009. doi: 10.1109/CIMSA.2009.5069941
72. Chr. von der Malsburg, “Self-organization of Orientation Sensitive Cells in the Striate
Cortex”, Journal of High Energy Physics, vol.4, no. 2, pp, 85–100, Jun. 1973. doi:
10.1007/BF00288907
73. Elsawy, Hesham, E. Hossain and I. K. Dong, “HetNets with Cognitive Small Cells: User
Offloading and Distributed Channel Access Techniques”, IEEE Communications Magazine,
vol. 51, no. 6, pp. 28–36, Jun. 2013. doi: 10.1109/MCOM.2013.6525592
74. Mangiameli, P., Chen, S. K., & West, D. (1996), “A Comparison of SOM Neural Network
and Hierarchical Clustering Methods”, European Journal of Operational Research, vol. 93,
no. 2, pp. 402–417, Sep. 6th , 1996. doi:10.1016/0377-2217(96)00038-0
75. Owsley, L., Atlas, L., and Bernard, G. (1996), “Self-Organizing Feature Maps with Perfect
Organization”, International Conference on Acoustics, vol. 6, pp. 3557–3560, May. 7th -10th ,
1996. doi: 10.1109/ICASSP.1996.550797
154 4 Intelligent Network Control

76. Chip-Hong C, Pengfei X, Rui X, et al, “New Adaptive Color Quantization Method Based
on Self-Organizing Maps” [J], IEEE Transactions on Neural Networks, vol. 16, no. 1, pp.
237–249, Feb. 2005. doi: 10.1109/TNN.2004.836543
77. Jian Xianzhong, Cao Shujian, Guo Qiang, “Segmentation of CAPTCHA characters based on
self organizing maps and Voronoi”, Application Research of Computers, Vol.32, No.9, pp.
Sep.2015. doi: 2015, 32(9):2857–2861.
78. 3GPP TS 32.500, “Telecommunication Management; Self-Organizing Networks (SON);
Concepts and requirements”, Jul. 2008.
79. Gao Y, Wei Y, Fu D, et al, “Research and Application of a New Artificial Immune
Algorithm Which Based on SOM Neural Network”, 2006 IEEE International Conference on
Networking, Sensing and Control, pp. 1080–1083, 2006. doi: 10.1109/ICNSC.2006.1673302
80. Valero, S., Aparicio, J., Senabre, C., Ortiz, M., Sancho, J., and Gabaldon, A., “Comparative
Analysis of Self Organizing Maps vs. Multilayer Perceptron Neural Networks for Short-
Term Load Forecasting”, Modern Electric Power Systems (MEPS), 2010 Proceedings of the
International Symposium, pp. 1–5, Sept. 20th – 22nd , 2010.
81. Fang, C., Yu, F. R., Huang, T., & Liu, J., “A Distributed Energy-Efficient Algorithm in Green
Content-Centric Networks”, IEEE International Conference on Communications(IOC), pp.
5546–5551, June 8th -12th , 2015. doi: 10.1109/ICC.2015.7249206
82. Xiaodong X U, Zhang H, Dai X, et al, “SDN Based Next Generation Mobile Network
with Service Slicing and Trials” [J], Wireless Communication over Zigbee for Automotive
Inclination Measurement China Communications, vol. 11, no. 2, pp. 65–77, Feb. 2014. doi:
10.1109/CC.2014.6821738
83. Hai Z, Ding L and Xiang L, “Networking Scientific Resources in the Knowledge Grid
Environment: Research Articles” [J], Concurrency & Computation Practice & Experience,
vol. 19, no, 7, pp. 1087–1113, May. 2007. doi: 10.1002/cpe.1094
84. Chao Fang, F. Richard Yu, Tao Huang, Jiang Liu, and YunJie Liu, “Energy-Efficient
Distributed In-Network Caching for Content-Centric Networks”, Global Internet Symposium,
pp. 91–96, 2014.
85. ZHOU Kaili and KANG Yaohong, “Neural Network Model and its MATLAB Simulation Pro-
gram Design (Chinese Edition)”, Tsinghua University Press, 2005. ISBN: 9787302108290
86. Z Wang, X Wang, L Liu and M Huang, “Optimal State Feedback Control for Wireless
Networked Control Systems with Decentralized Controllers”, Let Control Theory and
Applications, vol. 9, no. 6, pp. 852–862, Apr. 2015. doi: 10.1049/iet-cta.2014.0418
87. Yang C T, Liu J C, Ranjan R, et al, “On Construction of Heuristic QoS Bandwidth
Management in Clouds” [J], Concurrency & Computation Practice & Experience, vol. 15,
no. 18, pp. 2540–2560, Dec. 2013. doi: 10.1002/cpe.3090
88. H Casanova, J Dongarra and DM Doolin, “Java Access to Numerical Libraries”, Concurrency
Practice & Experience, vol. 9, no. 11, pp. 1279–1291, Nov. 1997. doi:10.1002/(SICI)1096-
9128(199711)9:11<1279::AID-CPE339>3.0.CO;2-E
89. Chao Fang, F. Richard Yu, Senior Member, IEEE, Tao Huang, Jiang Liu, and Yunjie Liu,
“A Survey of Green Information-Centric Networking: Research Issues and Challenges”,
IEEE Communication Surveys & Tutorials, vol. 17, no. 3, pp. 1455–1472, July 2015. doi:
10.1109/COMST.2015.2394307
90. Chao Fang, F. Richard Yu, Tao Huang, Jiang Liu, and Yunjie Liu, “A Survey of Energy-
Efficient Caching in Information-Centric Networking”, IEEE Communications Magazine,
vol. 52, no. 11, pp. 122–129, Nov. 2014. doi: 10.1109/MCOM.2014.6957152
91. J.-Q. Li, F. R. Yu, G. Deng, C. Luo, Z. Ming, and Q. Yan, “Industrial internet: A survey on
the enabling technologies, applications, and challenges,” IEEE Comm. Surveys & Tutorials,
vol. 19, no. 3, pp. 1504–1526, 2017.
92. H. Zhang, C. Jiang, X. Mao, and H. H. Chen, “Interference-limited resource optimization in
cognitive femtocells with fairness and imperfect spectrum sensing,” IEEE Transactions on
Vehicular Technology, vol. 65, no. 3, pp. 1761–1771, 2016.
93. L. Cui, F. R. Yu, and Q. Yan, “When big data meets software-defined networking: SDN for
big data and big data for SDN,” IEEE Net., vol. 30, no. 1, pp. 58–65, 2016.
References 155

94. J. Wan, S. Tang, Z. Shu, D. Li, S. Wang, M. Imran, and A. V. Vasilakos, “Software-defined
industrial internet of things in the context of industry 4.0,” IEEE Sensors Journal, vol. 16,
no. 20, pp. 7373–7380, 2016.
95. H. R. Faragardi, H. Fotohi, T. Nolte, and R. Rahmani, “A cost efficient design of a multi-
sink multi-controller WSN in a smart factory,” in Proc. Conf. High Performance Comp. and
Comm., 2017.
96. X. Lei, C. Jiang, C. Yan, W. Jian, and R. Yong, “A framework for categorizing and applying
privacy-preservation techniques in big data mining,” Computer, vol. 49, no. 2, pp. 54–62,
2016.
97. K. Kaur, S. Garg, G. S. Aujla, N. Kumar, J. J. Rodrigues, and M. Guizani, “Edge computing
in the industrial internet of things environment: Software-defined-networks-based edge-cloud
interplay,” IEEE Comm. Mag., vol. 56, no. 2, pp. 44–51, 2018.
98. N. G. Nayak, F. Dürr, and K. Rothermel, “Incremental flow scheduling and routing in time-
sensitive software-defined networks,” IEEE Trans. on Industrial Informatics, vol. 14, no. 5,
pp. 2066–2075, 2018.
99. J. Wang, C. Jiang, Z. Han, Y. Ren, and L. Hanzo, “Network association strategies for an
energy harvesting aided super-wifi network relying on measured solar activity.” IEEE Journal
on Selected Areas in Comm., vol. 34, no. 12, pp. 3785–3797, 2016.
100. S. Nakamoto, “Bitcoin: A peer-to-peer electronic cash system,” http://bitcoin.org/bitcoin.pdf/,
Last Accessed Aug. 2018.
101. F. R. Yu, J. Liu, Y. He, P. Si, and Y. Zhang, “Virtualization for distributed ledger technology
(vDLT),” IEEE Access, vol. 6, pp. 25 019–25 028, 2018.
102. Y. Shen, C. Jiang, T. Quek, and R. Yong, “Device-to-device assisted communication in
cellular network—an energy efficient approach in downlink video sharing scenario,” IEEE
Transactions on Wireless Communications, vol. 15, no. 2, pp. 1575–1587, 2016.
103. N. Group, “The difference between permissionless and permissioned networks,” https://
medium.com/netis-group-blog/, Last Accessed Aug. 2018.
104. G. Wood, “Ethereum: A secure decentralised generalised transaction ledger,” Ethereum
Project Yellow Paper, vol. 151, pp. 1–32, 2014.
105. C. Cachin and M. Vukolić, “Blockchains consensus protocols in the wild,” arXiv preprint
arXiv:1707.01873, 2017.
106. Y. H. Yang, Y. Chen, C. Jiang, and K. J. R. Liu, “Wireless network association game with
data-driven statistical modeling,” IEEE Transactions on Wireless Communications, vol. 15,
no. 1, pp. 512–524, 2016.
107. M. Castro and B. Liskov, “Practical byzantine fault tolerance and proactive recovery,” ACM
Trans. on Comp. Sys., vol. 20, no. 4, pp. 398–461, 2002.
108. H. Yao, P. Si, R. Yang, and Y. Zhang, “Dynamic spectrum management with movement
prediction in vehicular ad hoc networks.” Adhoc & Sensor Wireless Networks, vol. 32, no. 11,
2016.
109. L. Lamport et al., “Paxos made simple,” ACM Sigact News, vol. 32, no. 4, pp. 18–25, 2001.
110. C. Cachin, “Architecture of the Hyperledger blockchain fabric,” in Workshop on Distributed
Cryptocurrencies and Consensus Ledgers, 2016, pp. 121–125.
111. J. Guo, X. Liu, C. Jiang, J. Cao, and R. Yong, “Distributed fault-tolerant topology control in
cooperative wireless ad hoc networks,” IEEE Transactions on Parallel & Distributed Systems,
vol. 26, no. 10, pp. 2699–2710, 2015.
112. C. Jiang, Y. Chen, Y.-H. Yang, C.-Y. Wang, and K. R. Liu, “Dynamic chinese restaurant game:
Theory and application to cognitive radio networks,” IEEE Trans. on Wireless Comm., vol. 13,
no. 4, pp. 1960–1973, 2014.
113. J. Wang, C. Jiang, H. Zhang, X. Zhang, V. C. Leung, and L. Hanzo, “Learning-aided network
association for hybrid indoor LiFi-WiFi systems,” IEEE Trans. on Veh. Tech., vol. 67, no. 4,
pp. 3561–3574, 2018.
114. C. Jiang, Y. Chen, Y. Gao, and K. R. Liu, “Indian buffet game with negative network
externality and non-bayesian social learning,” IEEE Trans. on Sys., Man, and Cybernetics:
Sys., vol. 45, no. 4, pp. 609–623, 2015.
156 4 Intelligent Network Control

115. C. Qiu, F. R. Yu, F. Xu, H. Yao, and C. Zhao, “Permissioned blockchain-based distributed
software-defined industrial internet of things,” in Globecom Workshops (GC Wkshps), 2018
IEEE, 2018, pp. 1–7.
116. J. Du, C. Jiang, G. Qiang, M. Guizani, and R. Yong, “Cooperative earth observation through
complex space information networks,” IEEE Wireless Communications, vol. 23, no. 2, pp.
136–144, 2016.
117. C. Qiu, S. Cui, H. Yao, F. Xu, F. R. Yu, and C. Zhao, “A novel QoS-enabled load scheduling
algorithm based on reinforcement learning in software-defined energy internet,” Future
Generation Comp. Sys., 2018.
118. C. Qiu, C. Zhao, F. Xu, and T. Yang, “Sleeping mode of multi-controller in green software-
defined networking,” EURASIP Journal on Wireless Commu. and Net., vol. 2016, no. 1, p.
282, 2016.
119. H. Yao, C. Qiu, C. Zhao, and L. Shi, “A multicontroller load balancing approach in software-
defined wireless networks,” International Journal of Distributed Sensor Net., vol. 11, no. 10,
p. 454159, 2015.
120. P.-L. Aublin, S. B. Mokhtar, and V. Quéma, “RBFT: Redundant byzantine fault tolerance,” in
Proc. Conf. Distributed Comp. Sys.’ 13, 2013, pp. 297–306.
121. A. Clement, E. L. Wong, L. Alvisi, M. Dahlin, and M. Marchetti, “Making byzantine fault
tolerant systems tolerate byzantine faults.” in NSDI, vol. 9, 2009, pp. 153–168.
122. “Hyperledger indy,” https://cn.hyperledger.org/projects/hyperledger-indy, Last Accessed
Aug. 2018.
123. C. J. Watkins and P. Dayan, “Q-learning,” Machine learning, vol. 8, no. 3–4, pp. 279–292,
1992.
124. V. Mnih, K. Kavukcuoglu, D. Silver, A. A. Rusu, J. Veness, M. G. Bellemare, A. Graves,
M. Riedmiller, A. K. Fidjeland, G. Ostrovski et al., “Human-level control through deep
reinforcement learning,” Nature, vol. 518, no. 7540, p. 529, 2015.
125. Z. Wang, T. Schaul, M. Hessel, H. Van Hasselt, M. Lanctot, and N. De Freitas, “Dueling
network architectures for deep reinforcement learning,” arXiv preprint arXiv:1511.06581,
2015.
126. M. Abadi, A. Agarwal, P. Barham, E. Brevdo, Z. Chen, C. Citro, G. S. Corrado, A. Davis,
J. Dean, M. Devin et al., “Tensorflow: Large-scale machine learning on heterogeneous
distributed systems,” arXiv preprint arXiv:1603.04467, 2016.
127. T. M. Chilimbi, Y. Suzue, J. Apacible, and K. Kalyanaraman, “Project Adam: Building an
efficient and scalable deep learning training system.” in OSDI, vol. 14, 2014, pp. 571–582.
Chapter 5
Intelligent Network Resource
Management

Resource management problems are ubiquitous in the networking field, such as job
scheduling, bitrate adaptation in video streaming and virtual machine placement
in cloud computing. In this chapter, we propose a reinforcement learning based
dynamic attribute matrix representation (RDAM) algorithm for virtual network
embedding. The RDAM algorithm decomposes the process of node mapping into
the following three steps: (1) static representation of substrate physical network. (2)
dynamic update of substrate physical network. (3) Reinforcement-Learning-Based
algorithm. Then, we design and implement a policy network based on reinforcement
learning to make node mapping decisions. We use policy gradient to achieve
optimization automatically by training the policy network with the historical data
based on virtual network requests.

5.1 Virtual Network Embedding Based on RDAM

Network virtualization [1, 3] is regarded as an essential technology for the new

generation of Internet architectures. Using network virtualization technology, ISP
can create multiple virtual networks on the same substrate physical network,
providing users with a variety of customizable end-to-end services [10]. In order
to make use of substrate physical network resources effectively, it is necessary to
utilize a virtual network embedding algorithm. Virtual network embedding maps
virtual requests to specific nodes and links in the substrate physical network. The
goal of the virtual network embedding is to make full use of the substrate physical
network resources with certain resource constraints.
The virtual network requests contain some constraints, such as the constraints
of the attributes of node links, the constraints of request admission, the constraints
of dynamically changing request, and so on [6]. Considering multiple constraints
simultaneously will make it tough for the virtual network embedding problem

© Springer Nature Switzerland AG 2019 157

H. Yao et al., Developing Networks using Artificial Intelligence,
Wireless Networks, https://doi.org/10.1007/978-3-030-15028-0_5
158 5 Intelligent Network Resource Management

to obtain the optimal node. The optimization process is NP-hard [2, 4, 18, 36].
Therefore, the majority of virtual network embedding algorithms heuristically add
assumptions and constraints to reduce the solution space. In this way, the optimal
solution can be obtained within an acceptable complexity.
However, the results based on a series of rules and assumptions are not particu-
larly convincing [26, 34]. In addition, the heuristics based on hand-crafted rules are
not universal for multiple evaluation metrics. The RDAM algorithm introduced in
this section is different from traditional method in three perspectives.
1. Static representation of substrate physical network. The nodes and links infor-
mation of substrate physical network can be represented by the attribute matrix
and the adjacency matrix, respectively. However, these two kind of matrices may
be incomplete and noisy. We obtain a robust consensus matrix through spectrum
method, and this consensus matrix can effectively represent the substrate physical
network.
2. Dynamic update of the substrate network. Substrate physical network needs to be
updated after virtual network request arrives. In the virtual network embedding
problem, available physical network is changing dynamically at a high frequency
[5, 7, 40]. If the consensus matrix is updated every time with the spectrum
method, the computational complexity cannot be satisfied in real scenario. We
utilize the perturbation theory to capture the changes of nodes and links in the
substrate network under continuous time and to complete an efficient method of
updating substrate physical network.
3. Reinforcement learning based algorithm. The reinforcement learning agent can
effectively discover the relationship between the substrate network representation
and virtual network requests, thereby completing an efficient virtual network
embedding [11].

5.1.1 System Model and Problem Formulation

5.1.1.1 Description of Virtual Network Embedding

Figuer 5.1 shows the mapping process of two different virtual network requests,
where Fig. 5.1a and b are two different virtual requests, and Fig. 5.1c is the substrate
network. An undirected graph GS = (N S , LS , ASN , ASL ) is applied to denote
the substrate network, where N S and LS denote the set of substrate nodes and
links, respectively. ASN and ASL denote the attributes of substrate nodes and links,
respectively. Similarly, virtual request can be represented by an undirected graph
GV = (N V , LV , CN V , C V ), where N V and LV denote the set of substrate nodes and
L
links, respectively. CNV and C V denote the constraints of substrate nodes and links,
L
respectively.
We take Fig. 5.1a as an example. It can be seen from squares boxes that virtual
node r1 requires 5 units of computing resources and virtual node r2 requires 10 units
5.1 Virtual Network Embedding Based on RDAM 159

5 10 30 40 10
10 20 10
r1 r2 s1 s2 s3
Request1 30
(a)

10 30 s4 10
r3
20
10 20

10
r4 r5 s5 30
s6 20
s7
5 10 20 10 20
Request2 Substrate Network
(b) (c)

Fig. 5.1 Virtual network request and substrate network

of computing resources. The virtual link between virtual node r1 and virtual node
r2 requires 10 units of link resources. Figure 5.1c shows a substrate network that
contains 7 substrate nodes from node s1 to node s7. The substrate node s1 holds 30
units of computing resources and the substrate node s2 holds 40 units of computing
resources. The link resources between node s1 and node s2 reach 20 units.
The virtual network embedding process can be formulated as a mapping M :

GV (N V , LV ) → GS (N , L ), where N ⊂ N S and L ⊂ LS . As shown in Fig. 5.1,
virtual node r1 and virtual node r2 may be mapped to substrate node s1 and substrate
node s2, respectively. Then the link request between r1 and r2 is mapped to substrate
link between s1 and s2. Note that it is also possible that r1 and r2 are mapped to
substrate node s1 and substrate node s3, respectively. Then the link request between
r1 and r2 is mapped to two substrate links between substrate node s1 and substrate
node s3. Obviously, the latter is not optimal because the link resource between
substrate node s2 and substrate node s3 is consumed redundantly.
We use t denote the arrival time of the virtual request in Fig. 5.1a, td denote the
duration of the virtual request. In the period of td , the substrate resources that have
been allocated to a request cannot be allocated to other virtual request. Therefore,
the virtual network embedding algorithm will have a significant impact on the
utilization of substrate resources.

5.1.1.2 Metrics of Virtual Network Embedding

The main goal of virtual network embedding is to map virtual network requests to
substrate networks as many as possible [9]. It is beneficial to increase the utilization
160 5 Intelligent Network Resource Management

efficiency of substrate network resources and increase the revenue of infrastructure

operators. Most works use three metrics, i.e., long-term average revenue, long-term
revenue to cost ratio and long-term acceptance ratio, to measure the utilization
efficiency of substrate network resources.
For brevity, in the rest of this part, computing resource is represented by CPU,
and link resource is represented by bandwidth. According to the description in [50],
the revenue of a virtual network request is

R(GV , t, td ) =
⎡ ⎤
  (5.1)
td ⎣wc CP U (nV ) + wb BW (LV )⎦
nV ∈N V l V ∈LV

where wc and wb are the weights for CPU and bandwidth, respectively. When a
virtual request is allocated, the revenue is generated. No revenue is generated if the
request is rejected. The virtual request lasting longer have more revenue.
We define the cost of accepting a virtual request as follows

C(GV , t, td ) =
⎡ ⎤
   (5.2)
td ⎣ BW (fllS )⎦
V
CP U (nV ) +
nV ∈N V l V ∈LV l S ∈LS

V
where fllS denotes the total bandwidth l S allocated to the link request l V by the
virtual network embedding algorithm. In the process of link mapping, a link request
l V may be allocated to multiple links, so the total bandwidth consumption needs to
be calculated.
The first metric is long-term average revenue, which is the ratio of revenue to
time in infinite time horizon. It measures the overall effect of a virtual network
embedding algorithm. The long-term average revenue is defined as
T V
t=0 R(G , t, td )
Rev = lim . (5.3)
T →∞ T

In Eq. (5.3), Rev is the long-term average revenue, Tt=0 R(GV , t, td ) is the total
revenue in infinite time horizon. Then it is necessary to achieve a high long-term
average revenue with the substrate network resources are consumed less, so we
define the long-term revenue to cost ratio
T
R(GV , t, td )
RevT oCos = lim Tt=0 . (5.4)
T →∞ V
t=0 C(G , t, td )
5.1 Virtual Network Embedding Based on RDAM 161


In Eq. (5.4), RevT oCos is the long-term revenue to cost ratio, Tt=0 C(GV , t, td )
is the total cost in infinite time horizon. Finally, We define the third evaluation metric
long-term acceptance ratio Accept, which is the ratio of the number of accepted
requests to the total number of virtual requests. Defining the long-term acceptance
rate contributes to increasing the probability that virtual network requests are
accepted, and reduce the number of rejected virtual network requests.

5.1.1.3 Markov Decision Process

We model the virtual network embedding problem as a sequence of decision making

instances where an agent receives virtual network requests [12–15]. The agent
solves the virtual node mapping and receives a reward. The agent’s objective is
to maximize this reward.
We use Markov Decision Process (MDP) to model the sequential decision-
making problem like [50]. A discrete time MDP M is a quintuple (S, A, R, P , γ ),
where S is the state space, A is the action space, R : S × A → R is the reward
function that assigns real-valued rewards to state-action pairs, and P : S × A × S →
[0, 1] is a transition probability distribution. Therefore, R(st = s, at = a) is the
reward for performing action a ∈ A in state st and

P (s , a, s) = P r(st+1 = s |at = a, st = s) (5.5)

is the probability of a transition to state s when selecting action a in state s. The

behavior of a decision-making agent is defined by its policy π of selecting actions.
The overall return of a given policy π is calculated by
∞

Rt = rt+1 + γ rt+2 + γ rt+3 . . . =
2
γ k rt+k+1 , (5.6)
k=0

where 0 ≤ γ t ≤ 1 is a discount factor, rt+1 represents the revenue in time step t+1.
The goal of the decision-making agent is to find the policy π ∗ that maximizes the
expected cumulative rewards given an initial state s1.

5.1.1.4 Feature Attribute Extraction

In the RDAM algorithm, the description of the substrate node attribute is very
important for the reinforcement learning agent to comprehend substrate network
and learn mapping rules. The work presented in [20] extracts the following substrate
node attributes:
1. Computing Resources (CP U ): The computing resource of a substrate node
determines its availability. When the computing resource of a substrate node is
sufficient, it is more likely to accept node requests.
162 5 Intelligent Network Resource Management

2. Degree (DEG): The degree denotes the connectivity of substrate network. Nodes
with better connectivity are more likely to be occupied by node requests.
3. Sum of bandwidth (SU M BW ): The sum of substrate network bandwidth
describes connectivity from the perspective of available bandwidth. When a
substrate node has access to more bandwidth, mapping a virtual node to it may
lead to better link mapping options.
4. Average distance to other host nodes (AV GDST ): The fourth attribute of a
substrate node is the average distance from it to other already mapped node in
the same request. If we select a substrate node close to those already mapped, the
cost of bandwidth can be effectively reduced. The combination of mapped nodes
with few hops can effectively save the link resources of the substrate network.
After above four attributes are extracted, they need to be normalized to facilitate
subsequent node sorting and optimization. The normalization method for CPU,
DEG, and SU M BW is to divide the corresponding attribute value by the maximum
attribute value in the initial state of substrate physical network. For AV GDST , the
initialization mode is
1
. (5.7)
AV G(DST ) +1

By concatenate these four attributes together, we get the node’s attribute matrix A,
i.e.,

A = [CP U, DEG, SU M (BW , AV G(DST ) ]. (5.8)

We summarize some notations in the RDAM algorithm. The uppercase symbol

represents a matrix (eg, A), the arrow symbol represents a vector (eg, a), and the
lower case symbol represents a scalar (eg, a). A denotes the transpose of the matrix
A. As we describe the physical network GS = (N S , LS , ASN , ASL ) in Sect. 5.1.1.1,
ASN represents the attributes of the nodes and ASL represents the attributes of the
links. Therefore, A(t) denotes an attribute matrix at time step t, and X(t) denotes an
adjacency matrix at time step t. In Table 5.1, the definitions of all the symbols used
in the RDAM algorithm are given.

5.1.2 RDAM Algorithm

This subsection will mainly introduce the RDAM algorithm, including static
representation of substrate network, dynamic update of substrate network, model
construction, optimization method and, the overall steps of training and testing.
The RDAM algorithm regards virtual network embedding as two stages: node
mapping and link mapping. In the link mapping stage, the nodes already mapped are
assigned links by breadth-first traversal. The process of node mapping is described
5.1 Virtual Network Embedding Based on RDAM 163

Table 5.1 Symbols

Notations Definitions
A(t) Attribute matrix at time step t
X(t) Adjacency matrix at time step t
A(t+1) Attribute matrix at time step t+1
X(t) Adjacency matrix at time step t+1
ΔA Change of attribute matrix between time steps t and t+1
ΔX Change of adjacency matrix between time steps t and t+1
n Number of nodes in substrate network
d Number of node attributes
k Embedding dimension for network attributes or structure
l Final consensus embedding dimension

in detail in Fig.5.2. It is divided into Problem 1 and Problem 2. Problem 1 refers

to the problem of static representation of substrate network. Problem 2 refers to the
problem of dynamic update of the substrate network.
We utilize the spectrum method in problem 1 to obtain a robust consensus matrix
of substrate network. We adopt the matrix perturbation theory in problem 2 to reduce
complexity of the static update method.

5.1.2.1 Static Representation of Substrate Network

The substrate network has node information and link information. The node
information of substrate network can be represented by an attribute matrix; the link
information of substrate network can be represented by an adjacency matrix. We
divide the process of static representation of substrate network into two steps. The
first step is to reduce the noise from the attribute matrix A(t) and adjacency matrix
(t)
X(t) . We obtain the embedding representation YA of the attribute matrix X(t) and
(t)
the embedding representation YX of A(t) . In the second step, the final consensus
(t) (t)
matrix Y (t) is obtained from YX and YA .
(t) (t) 
Let DX ∈ R n×n be the degree matrix of X(t) , i.e., DX (i, i) = nj=1 X(t) (i, j ).
(t) (t)
Then LX = DX − X(t) is a Laplacian matrix. According to spectral theory
[21, 22], mapping an n-dimensional matrix to a k-dimensional embedding matrix
(k  n) can effectively reduce the noise in the matrix representation. A universal
choice of YX = [y1 , y2 , . . . , yn ] ∈ Rn×k is to minimize the loss function
(t)
 n
j =1 X (i, j )||yi − yj ||2 . In the embedding space, the distance between
1 (t) 2
2
interconnected nodes is closer.
The first step degenerates into a generalized eigen-problem L(t) (t)
X a = λDX a [23].
Assuming a1 , a2 , . . . , an are the eigenvectors of the corresponding eigenvalues 0 =
λ1  λ2  . . .  λn , it is easy to verify λ1 = 0 and the corresponding eigenvector is
164 5 Intelligent Network Resource Management

t t+1 Time

f1 f2 f3 f4 f5 f1 f2 f3 f4 f5 f1 f2 f3 f4 f5 f1 f2 f3 f4 f5

A B A B
Substrate Substrate
Network Network

D C D C
f1 f2 f3 f4 f5 f1 f2 f3 f4 f5 f1 f2 f3 f4 f5 f1 f2 f3 f4 f5

Aributes Network
Aributes Network
Changes Changes

f1 f2 f3 f4 f5 A B C D f1 f2 f3 f4 f5 A B C D
A A A A
B B B B
C C C C
D D D D
A X ΔA ΔX
Online Online
Embedding Embedding
Update Update

Ya Yx New Ya New Yx

Consensus Consensus
Embedding Embedding

Y Problem1 New Y Problem2

Fig. 5.2 The left side denotes the problem 1 at time step t and the right side denotes the problem 2
at time step t+1. At time step t, we first obtained the attribute matrix A(t) and adjacency matrix X(t)
of substrate network. The noise in aforementioned two matrices is eliminated by spectral analysis.
We obtain the embedding representation YA(t) and YX(t) of the attribute matrix and adjacency matrix.
Then we get the final consensus matrix through two embedding matrices. Then at time step t+1,
the grey grid in substrate network denotes the changed attribute ΔA and changed structure ΔX.
(t) (t)
We will obtain the new embedding representation YA and YX from the variation of the attribute
matrix and the adjacency matrix. Finally, we get the new consensus matrix Y (t+1)

unit vector 1. Then the k-dimensional embedding YX(t) ∈ Rn×k of network structure
(t)
is given by the top-k eigenvectors starting from a2 , i.e., YX = [a2 , a3 , . . . , ak+1 ].
(t)
Similar to the adjacency matrix, the attribute embedding matrix YA can also
be obtained in such a way. First we normalize the attribute matrix and obtain
the similarity matrix W (t) of the normalized attribute matrix. Then we solve the
generalized eigen-problem and obtain the embedding representation of the attribute
(t)
matrix YA .
5.1 Virtual Network Embedding Based on RDAM 165

The noise of X(t) and A(t) has been eliminated by computing adjacency embed-
ding matrix YX(t) and the attribute embedding matrix YA(t) . Now we utilize them to
seek a consensus matrix. However, these two embedding matrices are obtained in
different ways and may not be relevant. In order to capture their interdependency
and make them compensate each other, we maximize the correlation of them. We
utilize two projection vectors PX(t) and PA(t) to maximize the correlation of YX(t) and
(t)
YA after projection. The problem is equivalent to

(t) (t) (t) (t) (t) (t) (t) (t)

max PA YA YA PA + PA YA YX PX
(t) (t)
PA ,PX

(t) (t) (t) (t) (t) (t) (t) (t) (5.9)

+ PX YX YA PA + PX YX YX PX ,

s.t. PA(t) YA(t) YA(t) PA(t) + PX(t) YX(t) YX(t) PX(t) = 1.

This is a typical optimization problem. Let γ be a Lagrangian operator. Then

(t) (t)
the value of PX and PA can be obtained by setting the derivative of the
(t) (t)
Lagrangian function w.r.t. PX and PA to zero. The expression of the corresponding
generalized eigen-problem is

     
(t) (t) (t) (t) (t) (t) (t) (t)
YA YA YA YX PA YA YA 0 PA
(t) (t) (t) (t) =γ (t) (t) .
YX YA YX YX PX(t) 0 YX YX PX(t)
(5.10)

We take the top-l eigenvectors as the consensus matrix of the above generalized
eigen-problem. Assuming a projection matrix P (t) ∈ R2k×l , then the final consensus
(t) (t)
matrix is expressed as Y (t) = [YA , YX ] × P (t) ∈ Rn×l .

5.1.3 Dynamic Update of Substrate Network

In the virtual network embedding process, the substrate network will change when
a virtual request is accepted. Therefore, the attribute matrix and adjacency matrix
are changing frequently. If we obtained the consensus matrix Y (t) by computing
eigenvectors and eigenvalues of generalized eigen-problem at each time step, this
method will be time-consuming and it is not practical in large-scale networks.
(t) (t)
Therefore, we propose to update YA and YX by matrix perturbation theory, which
can reduce the computational complexity.
We propose the dynamical update method based on the fact that the attribute
matrix and adjacency matrix do not change too much in two consecutive time
steps, as virtual request resources usually do not occupy a large part of substrate
166 5 Intelligent Network Resource Management

resources. We utilize ΔA and ΔX to donate the variation of the attribute matrix and
the adjacency matrix within two consecutive time steps. The degree matrix and the
Laplacian matrix are given by

(t+1) (t+1) (t+1) (t+1)

DA = DA + ΔDA , LA = LA + ΔLA ,
(5.11)
(t+1) (t+1) (t+1) (t+1)
DX = DX + ΔDX , LX = LX + ΔLX .

We take the update method of the embedding matrix of the adjacency matrix as
an example to illustrate the specific steps of dynamic update method. According to
matrix perturbation theory [24], we have

(L(t)
X + ΔLX )(a + Δa) =
(5.12)
(t)
(λ + Δλ)(DX + ΔDX )(a + Δa).

Then, for a certain eigenvalue and eigenvector pair (λi , ai ), we have

(t)
(LX + ΔLX )(ai + Δai ) =
(5.13)
(t)
(λi + Δλi )(DX + ΔDX )(ai + Δai ).

The algorithm problem of dynamic update is equivalent to how to obtain the

changed pairs (Δλi , Δai ) from the changed Laplacian matrix ΔL and the changed
degree matrix ΔD. The way to solve (Δλi , Δai ) is divided into following two parts.
Calculate Δλi
By expanding Eq. (5.13), we have

(t) (t)
LX ai + ΔLX ai + LX Δai + ΔLX Δai
(t) (t)
= λi DX ai + λi ΔDX ai + Δλi DX ai + Δλi ΔDX ai (5.14)
(t) (t)
+ (λi DX + λi ΔDX + Δλi DX + Δλi ΔDX )Δai ,

(t)
where these higher order terms, i.e., Δλi ΔDX ai , λi ΔDX Δai , Δλi DX Δai , and
Δλi ΔDX Δai can be removed as they has narrow effects on the accuracy of a
(t) (t)
generalized eigen-problem. In addition, since LX a = λDX a, Eq. (5.14) can be
simplified as

(t)
ΔLX ai + LX Δai
(5.15)
(t) (t)
= λi ΔDX ai + Δλi DX ai + λi DX Δai .
5.1 Virtual Network Embedding Based on RDAM 167

Next, multiplying both sides of Eq. (5.15) with ai , we have

ai ΔLX ai + ai LX Δai
(t)
(5.16)
= ai λi ΔDX ai + ai Δλi DX ai + ai λi DX Δai
(t) (t)

Since the Laplacian matrix L(t) (t)

X and the degree matrix DX are symmetric, we have

ai LX Δai = ai λi DX Δai
(t) (t)
(5.17)

So, Eq. (5.12) becomes

ai ΔLX ai = ai λi ΔDX ai + ai Δλi DX

(t)
ai . (5.18)

Finally, the change of eigenvalue Δλi is given by

ai ΔLX ai − ai λi ΔDX ai
Δλi = . (5.19)
ai DX ai
(t)

Theorem 5.1 In the generalized eigen-problem Av = λBv, if A and B are both

Hermitian matrices and B is positive-semidefinite matrice, the eigenvectors are real,
eigenvectors vj (i = j ) and B is orthogonal, then, vi Bvj = 0 and vi Bvi = 1.

In virtual network embedding problem: ai DX

(t)
aj = 0(i = j ) and ai DX
(t)
ai =
1(i = j )
(t) (t)
Proof Degree matrix DX and Laplacian matrix LX are symmetric Hermitian
(t)
matrices. At the same time, the degree matrix DX is a semi-positive definite matrix,
thus completing the proof. From above proof, we have

Δλi = ai ΔLX ai − ai λi ΔDX ai . (5.20)

Calculate Δai
Since the network structure changes smoothly in continuous time steps, we assume
that theperturbation of the eigenvectors consists of the top-k eigenvectors, i.e.,
Δai = k+1 j =2 αij aj , where αij is the weight of the j -th eigenvector.

By plugging Δai = k+1 j =2 αij aj into Eq. (5.15), we have

(t)

k+1
ΔLX ai + DX αij λj aj
j =2
(5.21)
(t) (t)

k+1
= λi ΔDX ai + Δλi DX ai + λi DX αij aj .
j =2
168 5 Intelligent Network Resource Management

When p = i, by multiplying both sides with ap (2  p  k + 1, p = i) on both

sides of Eq. (5.21) and using the theorem 1, we have:


k+1
ap ΔLX ai + ap DX
(t)
αij λj aj
j =2


k+1 (5.22)
= λi ap ΔDX ai + Δλi ap DX ai + λi ap DX
(t) (t)
αij aj
j =2

→ ap ΔLX ai + αip λp = λi ap ΔDX ai + αip λi .

Therefore, we obtain the weight αip as

ap ΔLX ai − λi ap ΔDX ai

αip = . (5.23)
λi − λp

When p = i, from Theorem 5.1, we have

(ai + Δai ) (DX + ΔDX )(ai + Δai ) = 1. (5.24)

By ignoring the high order term in Eq. (5.24), we get

2a i DX Δai + a i ΔDX ai = 0,
(t) (t)
(5.25)

namely,

1
αii = − a i ΔDX ai .
(t)
(5.26)
2
Therefore, the change of eigenvalue Δai is

1
Δai = − a i ΔDX ai ai
(t)
2
(5.27)

k+1
ap ΔLX ai − λi ap ΔDX ai
+ ( )aj .
λi − λp
j =2,j =i

Now, we get the perturbation pairs (Δλi , Δai ). The pseudo code is given in
Algorithm 1. At start time(t=1), we obtain the initial embedding attribute matrix
and embedding adjacency matrix by spectral method. In addition, we have the initial
eigenvalue and eigenvector pairs (λi , ai ) at start time. The input of the algorithm
is (λi , ai ) at start time.The algorithm computes the perturbation of the Laplacian
and degree matrices for each time step. The output is the first k eigenvalues and
eigenvector pairs at time step T .
5.1 Virtual Network Embedding Based on RDAM 169

In Algorithm 1, lines 3 and 4 call Eq. (5.20) and (5.27) to calculate the
perturbation, and then update the eigenvalues and eigenvectors. The embedding
matrix of attribute matrices can be updated in the same way.

Algorithm 1 Updating of embedding matrix for substrate network

Input: Top-k eigen-pairs of the generalized eigen-problem {(λ2 , − →
a2 ),(λ3 , −
→
a3 ),. . . ,(λk+1 , a−−→
k+1 )} at
start time(t=1). Perturbation of the diagonal matrix ΔLA and Laplacian matrix ΔDA at each
time.
Output: Top-k eigen-pairs {(λ2 , −
(T ) →(T )
a2 ), (λ3 , −
(T ) →(T )
a3 ),. . . , (λk+1 , −
a−→(T ) )} at time T.
(T )
k+1
1: for t = 1 → T − 1 do
2: for i = 2 → k + 1 do
3: Calculate the variation of Δλi by Eq. (5.20);
4: Calculate the variation of Δ−
→ai by Eq. (5.27);
(t+1) (t)
5: λi = λi + Δλi
6: −
→
ai (t+1) = −
→
ai (t) + Δ−→ai
7: end for
8: end for

5.1.3.1 Computational Complexity

The computational complexity of static method is O(n2 k + k 2 l), where n is number

of substrate nodes, k is the dimension of the embedding representation of the
adjacency or attribute matrix, and l is the dimension of the final consensus matrix.
Proof For an n × n square matrix, the computational complexity of solving
all the eigenvectors is O(n3 ), and the computational complexity of solving the
eigenvector corresponding to a certain eigenvalue is O(n2 ). In the static method, the
computational complexity of getting the embedding representation of the adjacency
matrix is O(n2 k), and the computational complexity of the embedding consensus
matrix is O(k 2 l).
The computational complexity of dynamic update is O(k 2 (n + l + la +
lx + da + dx )), where la , lx , da , dx are the number of non-zero elements in
ΔLA , ΔLX , ΔDA , ΔDX , respectively.
Proof In dynamic update method, the computational complexity of updating the
attribute embedding matrix and adjacency embedding matrix is O(k(la + da )) and
O(k(lx + dx )) respectively. Updating the eigenvectors of the attribute embedding
matrix and adjacency embedding matrix requires the computational complexity
O(k 2 (la +da +n)) and O(k 2 (lx +dx +n)). Finally, the computational complexity of
the embedding consensus matrix is O(k 2 l). So the total computational complexity
is O(k 2 (n + l + la + lx + da + dx )).
The attribute and adjacency matrix of a substrate physical network change
smoothly in continuous time steps Therefore, ΔLA , ΔLX , ΔDA , ΔDX are sparse,
170 5 Intelligent Network Resource Management

and the number of non-zero elements, la , lx , da , dx , will be very small. In addition,

according to spectrum method, k  n. Therefore, the computational complexity
of dynamic update will be much lower than the computational complexity of static
update. It is proved that RDAM is a practical dynamic update algorithm.

5.1.4 Network Modelling

5.1.4.1 Model Architecture

This subsection introduces the model used in the RDAM algorithm. The input of
model shown in Fig. 5.3 is the embedding consensus matrix Y (t) . And we will obtain
the probability of the virtual node to each physical network node.
As shown in Fig. 5.3, the model consists of four layers. The first layer is the input
j
layer, where Yi (1  i  n, 1  j  l) represents the value of the j -th dimension
in the hidden layer representation of the i-th substrate node. The second layer is
the convolutional layer [25]. After the convolution layers, we could get a vector
representation of n-dimensional available physical nodes, i.e.,

hi = w · Yi + b, (5.28)

where hi is the output of the i-th convolution layer, w is the weight vector of the
convolution kernel, and b is the bias.
The third layer is the softmax layer. The output of the convolutional layer h
is passed to the softmax layer. We will obtain a probability vector that represents
the virtual request node selecting each substrate node. For each physical node, the
probability softmax is calculated as

e hi
pi =  h . (5.29)
je
j

Input layer Convolutional layer Softmax layer Candidates layer

l
Y 1
1
Y 1
2
Y13 ··· Y1 l–1
Y1 a1 P1 P1

Y21 Y22 Y23 ··· Y2l–1 Y2l a2 P2 P2

···

···

···

···

···

Yn1 Yn2 Yn3 ··· Ynl–1 Ynl an Pn Pn

Fig. 5.3 The model architecture of RDAM

5.1 Virtual Network Embedding Based on RDAM 171

Softmax is a generalization of logistic functions. It can obtain the probability of

each dimension of a high-dimensional vector. The probability of each dimension
ranges (0,1), and the sum of all dimensions probability is 1.
The fourth layer is candidate node layer. Through this model, each physical
node has its probability. However, in actual situation, some physical node does not
satisfy the virtual request node. Therefore, the layer will have a filter to remove the
unsatisfied nodes. In this case, we will select the physical node with the highest
probability after candidate node layer.

5.1.4.2 Reinforcement Learning Agent

Let us consider a substrate network with n nodes and m links. At an arbitrary time,
the virtual network embedding solver receives a virtual network request GV that
requires p virtual nodes and q virtual links. We define the MDP corresponding to
virtual node mapping of GV as a finite-horizon MDP MGV . The decision-making
agent consecutively selects p substrate nodes for embedding nodes of GV , yielding
to p decision-making instances at discrete times t. We assume that in a given state
stV , the agent tries to identify a substrate node nV ∈ N S for embedding the first
element nV ∈ N V .
The state of StV at the decision making instance t is defined as

StV = (NtV = Nt−1

V
\ {nVt−1 }, NtS , Nt−1
S
\ {nSt−1 }), (5.30)

where nSt−1 is the substrate node selected for embedding the virtual node nVt−1 in
the previous time step. In the initial state, no virtual node has been embedded and,
thus, all substrate nodes are available for embedding the first virtual node. Hence,
N1S = N S and N1V = N V .
The agent selects a node nS ∈ {NtS ∩ N S (nVt )} from the set of viable actions,

AVt = {ε} ∪ {(nVt , ns ) : ∀nS ∈ {NtS ∩ N S (nVt )} (5.31)

where ε denotes an arbitrary action that forces a transition to a new state. As the
result of selecting a substrate node nst for embedding the virtual node nVt , the agent
receives a reward.
The size of the MDP state space grows exponentially with the number of state
variables. Complexity of exact algorithms for solving MDP such as Q-learning is
polynomial in the size of the state space. Therefore, finding exact solutions for
MDPs with large number of state variables is intractable. We utilize policy gradient
to solve this MDP problem and we will illustrate policy gradient method for virtual
embedding problem in the following subsection.
172 5 Intelligent Network Resource Management

5.1.5 Training and Testing

In supervised learning, samples include features and labels. After dealing with
features, the model will obtain a predicted label. The distance from the real label
and the predicted label is used as loss function. Supervised learning is a process
continuously reducing the loss function value and making the predicted label closer
to the real label. There is no real label of the samples in reinforcement learning.
However, reinforcement learning [27] has an evaluation metric. After dealing with
features, reinforcement learning chooses a predicted label randomly. If the predicted
label yields a better evaluation metric value, it indicates that the direction of the
model prediction is correct. In this case the parameters of the model are encouraged
to be trained in this direction. If the evaluation metric value brought by the predicted
label is small or even negative, it means that the direction of the model prediction is
not correct. The direction of parameter training of the model needs to be adjusted.
In virtual network embedding process, we assume that the physical network has
n nodes and each node embedding matrix represents l dimensions. The embedding
matrix (n × l) of physical network is passed as input to the model. At this point,
we cannot directly select the physical network node with the highest probability
because the network model parameters are initialized randomly. If the node with the
highest probability is selected, the model is always biased. Therefore, we need to
find a balance between the exploration of a better solution and the exploitation of
the existing model. We randomly select the i-th node from the probability vector P
and construct a one-hot encoded vector. That is, the vector has n dimensions, only
the i-th element is 1, and the rest are 0. Then, the loss function can be written as

L(y, p) = − yi log(pi ), (5.32)
i

where yi and pi are the value of the randomly selected one-hot vector and the
predicted probability vector respectively. Then we utilize the gradient derivative
to train the model. When this randomly selected node can yield a large evaluation
metric value, the direction of the model training is more inclined to make similar
decisions. When this randomly selected node yields a small evaluation metric
value, the model parameter training is not encouraged to make similar decisions.
Therefore, we update the gradient by

g := α · r · g (5.33)

where α is the learning rate which controls the speed of the model training. When α
is too large, the training process may not converge and the global optimal solution
may be missed. When α is too small, the training process is too slow. We need to
choose an appropriate learning rate. The larger rewards will have a greater impact on
learning agents by multiplying rewards with gradients. In this case, the model can
5.1 Virtual Network Embedding Based on RDAM 173

be more inclined to make similar decisions. And decisions that get smaller rewards
or negative rewards will have a smaller impact for learning agents.
For a virtual network request, there are generally multiple virtual network request
nodes. After dealing with each virtual request node, the RDAM algorithm will stack
the gradient instead of directly applying in the model because the virtual network
request may fail. If the embedding task fails, the corresponding stacked gradients
will be cleared and the next virtual network request is processed.
After the number of virtual network requests reaches the number of batches, all
the gradients in the stack are applied to the model, and then the stack is cleared. The
reason for we utilize the batch gradient descent is that the gradient update requires
a lot of time. If the batch gradient is adopted, it will save much time. Secondly, the
batch gradient averages the gradient in the batch size and the training results are
more stable.
A complete virtual network embedding training process is shown in Algorithm 2.
Lines 7–10 are the process of node mapping, and lines 11–13 are the process of link
mapping. Line 28 illustrates when the mapping fails, it clears the gradient in the
stack and starts to train the next virtual network embedding request. Lines 21–23
update the batch-size gradients and clear the request counter.

Algorithm 2 Training process

Input: Number of epochs numEpoch; Learning rate α; Training set;
Output: Trained parameters in policy network;
1: Initialize all the parameters in policy network;
2: while iteration < numEpoch do
3: for req ∈ trainingSet do count=0;
4: for node ∈ req do
5: Mf =getFeatureMatrix();
6: p=policyNet.getOutput(Mf ); //Get the probability distribution from policy network
7: host=sample(p); //Sample from the probability distribution to choose a node as host
8: computeGradient(host);
9: end for
10: if isMapped(∀ node ∈ req) then
11: bfsLinkMap(req);
12: end if
13: if isMapped(∀ node ∈ req, ∀ link ∈ req) then
14: reward=revToCost(req); //Compute revenue to cost ratio
15: multiplyGradient(reward, α); //Compute the final gradients
16: else
17: clear the stacked gradients;
18: end if
19: ++counter;
20: if counter reach the batch size then
21: apply gradients to parameters;
22: counter=0;
23: end if
24: end for
25: ++iteration;
26: end while
174 5 Intelligent Network Resource Management

Algorithm 3 Testing process

Input: Testing set;
Output: Long-term average revenue, acceptance ratio, long-term revenue to cost ratio;
1: Initialize all the parameters in policy network;
2: for req ∈ testSet do
3: for node ∈ req do
4: Mf =getFeatureMatrix();
5: host=maxProbabilty(p); //Greedy strategy
6: end for
7: bfsLinkMap(req);
8: if isMapped(∀ node ∈ req, ∀ link ∈ req) then
9: signal(SUCCESS);
10: end if
11: end for

The pseudocode for the test stage is shown in Algorithm 3. In the testing stage,
the RDAM utilizes the greedy policy to select the substrate node from the largest
probability node for mapping.

5.1.5.1 A Simple Example for RDAM Algorithm

Considering the topologies of the virtual network and the substrate network as
illustrated in Fig. 5.2, we can obtain that the adjacency matrix X at initial time is
⎡ ⎤
0101
⎢1 0 1 0⎥
X=⎢ ⎥
⎣0 1 0 1⎦ . (5.34)

1010
We assume that the attribute matrix A at initial time is

⎡ ⎤
20 10 10 30
⎢10 10 10 20⎥
A=⎢
⎣20
⎥. (5.35)
10 30 10⎦
10 20 10 30

As we illustrated in Algorithm 1, the first step is to get the static representation

of substrate physical network using spectrum method. We set the dimension k of the
5.1 Virtual Network Embedding Based on RDAM 175

embedding matrix YX 3 and set the dimension l of the embedding matrix YA 2. We

can obtain the consensus matrix Y
⎡ ⎤
0.24 7.07
⎢2.42 1.18⎥
Y =⎢
⎣5.22
⎥. (5.36)
1.82⎦
2.91 6.91

The embedding matrix Y ∈ Rn×l of physical network is passed as input to the

model. Then we can train the model following the Algorithm 2. We utilize the
perturbation theory defined in Algorithm 1 to capture the changes of nodes and
links in the substrate network under continuous time. When the training process
ends, the RDAM utilizes the greedy policy to select the substrate node from the
largest probability node for mapping.

5.1.6 Experiments
5.1.6.1 Datasets

This part utilizes the GT-ITM tool to generate the substrate network topology which
is commonly used in virtual network embedding algorithms. Finally, we form a
substrate network with approximately 100 nodes and 500 links, which is the size
of a medium-sized ISP. The CPU resources of each substrate node are uniformly
distributed from 50 to 100 units, and the bandwidth resources of each substrate link
are uniformly distributed from 20 to 50 units.
Similarly, we generate some virtual network requests. Each request has 2–10
virtual nodes. The CPU demand is uniformly distributed from 0 to 50. These virtual
nodes are connected to each other with a probability of 0.5 forming an average of (n-
1)/4 virtual links. Bandwidth requirements for virtual links are uniformly distributed
from 0 to 50 units. The process of a virtual request is a Poisson process with an
average of 4 requests over 100 time units.
In order to verify the generalization ability of the algorithm, we constructed 2000
virtual network requests. The training set contains 100 substrate physical nodes and
the first 1000 virtual network requests. The test set contains 100 substrate physical
nodes and the last 1000 virtual network requests. First, the model is trained by the
training set. If the trained network parameters have improved in the training set, the
same network parameters are used to perform virtual network embedding on the test
set. In this way we can observe the generalization effect of the training model on the
test set.
176 5 Intelligent Network Resource Management

5.1.6.2 Experimental Settings

Four-layers network is constructed by the Tensorflow and the network is initialized

using the parameters by normal distribution. Some hyperparameter settings are
involved in the training process. For example, the batch size is set to 100 and the
learning rate is set to 0.005. In the representation process of the physical network,
the dimension k of the embedding matrix is set to 20, and the dimension l of the
consensus matrix is set to 4.

5.1.6.3 Training Results

The reinforcement training process is more difficult to converge than the supervised
learning process. It requires an interaction process with environment to continuously
perceive the state of the environment. The learning agent makes decisions, adopts
certain behavior, receives rewards from the environment, and then adjusts its
strategies according to the rewards. Especially in the problem of virtual network
embedding such an NP-hard, it takes a very long time to converge.
Figure 5.4 shows the change of long-term average revenue, long-term revenue
to cost ratio and long-term acceptance ratio in 100 epochs. At the beginning of
the training (0 <epoch< 20), all three evaluation metrics perform poorly because
the parameters of the model are randomly initialized. In the middle of the training
(20 <epoch< 80), the evaluation metrics values start to get better, because the
random selection of physical nodes in the training process allows the model to
explore all possibilities of selecting physical nodes. When randomly selected nodes
have large revenues, the gradient update of the model will be large. It donates that
the learning agent remembers the rewards of such decisions and produces similar
decisions in subsequent decisions. At the later stage of training (80 <epoch< 100),
the metric values begin to fluctuate around a certain value, because the model is still
exploring the possibility of physical node selection, but the training process at this
time has converged.
Figure 5.5 shows the decreasing trend of the cross-entropy loss function during
training. It can be seen that the value of the loss function is constantly decreasing,
which also proves the effectiveness of the model. In the last 20 epochs, the loss
function tends to be a constant value, which means that the training process has
converged.

5.1.6.4 Testing Results

In order to verify the generalization of the model, we conduct experiments on the test
set. The other three algorithm were selected. The first one is the baseline algorithm
[37]:

H (ns ) = CP U (nS ) BW (LS )
(5.37)
l S ∈L(ns )
5.1 Virtual Network Embedding Based on RDAM 177

0.71
1200
1180 0.70
Long-term average revenue

1160 0.69
1140

Accept Ratio
0.68
1120
1100 0.67
1080
0.66
1060
1040 0.65
1020
0.64
1000
0 20 40 60 80 100 0 20 40 60 80 100
(a) Epoch (b) Epoch

0.39

0.38
Long-term revenue/cost

0.37

0.36
0.35

0.34

0.33

0.32

0.31

0 20 40 60 80 100
(c) Epoch

Fig. 5.4 Performance on training set. (a) The change of long-term average revenue in 100 epochs.
(b) The change of long-term revenue to cost ratio in 100 epochs. (c) The change of long-term
acceptance ratio in 100 epochs

In the training process, the long-term revenue consumption ratio is used as an

evaluation metric. In each epoch, if the long-term average revenue is higher than
the current best long-term average revenue, then the best long-term average revenue
will update. In addition, if the current best long-tern average revenue updates, the
model with current parameter will be used to perform virtual network embedding
algorithm on the test set. Obviously, the best evaluation metric value will be updated
frequently and the last result is the global best result.
In the test phase, the start time of the virtual network request is 22 and the end
time is 30,000. Assuming the time unit is 1000, Fig. 5.6 shows the change of the
evaluation metric value in thirty time units. In Fig. 5.6a and b, the long-term average
revenue and acceptance ratio will decrease in the initial stage, because physical
resources will be gradually consumed as virtual network requests arriving. The
initial period of long-term revenue and consumption ratio will not be significantly
reduced, because it has nothing to do with the number of physical resources.
It can be observed that the RL algorithm has a higher accept ratio at the very
beginning. It may be because the RL algorithm allocates the top resources. As the
178 5 Intelligent Network Resource Management

Fig. 5.5 Loss on training set

three metrics tend to be stable, it can be observed that in the three evaluation metrics,
the convergence trend and convergence value of the RDAM algorithm are better than
the other three algorithms. We can conclude that the reinforcement learning agent
learns the relationship of the physical network nodes during the training stage, and
the model can be generalized during the testing stage.

5.2 Virtual Network Embedding Based on Policy Network

and Reinforcement Learning

The combination of network virtualization and software defined networks is consid-

ered as the foundation towards the next generation of Internet architecture [28, 30].
Network virtualization enables the coexistence of multiple heterogeneous virtual
networks on a shared network [19, 31–33, 35]. For Internet Service Providers (ISPs),
it enables new business models of hosting multiple concurrent network services on
their infrastructures. Decisions for embedding are challenging problems for ISPs
since it determines the effectiveness of utilization of network resources [16]. A sub-
optimal embedding algorithm will decrease the overall capacity of the infrastructure
and lead to cost of revenue for ISPs. A virtual network consists of several virtual
nodes (e.g. virtual routers), connected by a set of virtual links. The purpose of virtual
network embedding is to map virtual networks to a shared physical network while
providing the requests with adequate computing and bandwidth resources.
However, the virtual network embedding problem has been proved to be NP-
hard [36]. As a result, a large number of heuristic algorithms have been proposed
5.2 Virtual Network Embedding Based on Policy Network and Reinforcement. . . 179

a
2400
RDAM
Long-term average revenue rl
2200 NodeRank
Baseline
2000

1800

1600

1400

1200

0 5 10 15 20 25 30
Time
b
0.82
RDAM
0.80 rl
NodeRank
0.78 Baseline

0.76
Accept Ratio

0.74

0.72

0.70

0.68

0.66
0 5 10 15 20 25 30
Time
c
RDAM
0.54 rl
Long-term revenue/cost

NodeRank
0.51 Baseline

0.48

0.45

0.42

0.39

0.36

0 6 12 18 24 30
Time

Fig. 5.6 Performance on testing set. (a) The change of long-term average revenue in 30 time units.
(b) The change of long-term revenue to cost ratio in 30 time units. (c) The change of long-term
acceptance ratio in 30 time units
180 5 Intelligent Network Resource Management

[8, 37–39, 42], but most of them rely on artificial rules to rank nodes or make
mapping decisions. The parameters in these algorithms are always fixed and cannot
be optimized, making the embedding decisions sub-optimally. On the other hand,
in prior works, the information about substrate network and the knowledge about
virtual network embedding hidden in historical network request data have always
been overlooked. Historical network requests are a good representation of temporal
distribution and resource demands in the future.
In recent years, big data, machine learning and artificial intelligence have exciting
breakthroughs achieving state of the art results such as natural language understand-
ing and object detection [17]. Machine learning algorithms process a large amount
of data collected during a period and automatically learn the statistical information
from the data to give classification or prediction. Reinforcement learning, as a
widely-used technique in machine learning, has shown a great potential in dealing
with complex tasks, e.g., game of go [43], or complicated control tasks such as auto-
driving and video games [29, 45]. The goal of a reinforcement learning system (or
an agent) is to learn better policies for sequential decision making problems with an
optimal cumulative future reward signal [44, 46].
In this part, we introduce reinforcement learning into the problem of virtual
network embedding to optimize the node mapping process. Similar to earlier works
[8, 37, 48], our work is based on the assumption that all network requests follow
an invariable distribution. We divide our network request data into a training set
and a testing set, to train our Reinforcement Learning Agent (RLA) and evaluate
its performance respectively. We devise an artificial neural network called policy
network as the RLA, which observes the status of substrate network and outputs
node mapping results. We train the policy network with historical network request
data using policy gradient through back propagation. An exploration strategy is
applied in the training stage to find better solutions, and a greedy strategy is applied
in evaluation to fully evaluate the effectiveness of the RLA. Extensive simulations
show that the RLA is able to extract knowledge from historical data and generalize
it to incoming requests. To the best of our knowledge, this work is the first to utilize
historical network requests data and policy network based reinforcement learning
to optimize virtual network embedding automatically. The RLA outperforms two
representative embedding algorithms based on node ranking in terms of long-term
average revenue and acceptance ratio, while making a better utilization of network
resources [41].

5.2.1 Network Modelling

In this subsection, we present a network model and formulate the virtual network
embedding problem with description of its components. The notations used in this
section are shown in Table 5.2.
5.2 Virtual Network Embedding Based on Policy Network and Reinforcement. . . 181

Table 5.2 Frequently used GS Substrate network

notations
NS Nodes of substrate network
LS Links of substrate network
ASN Node attribute of substrate network
ASL Link attribute of substrate network
GV Virtual network of a certain virtual request
NV Nodes of a virtual network
LV Links of a virtual network
AVN Constraints of substrate nodes
AVL Constraints of substrate nodes

Fig. 5.7 An example of virtual network embedding

Figure 5.7 shows the mapping process of two different virtual network requests.
A substrate network is represented as an undirected graph GS = (N S , LS , ASN , ASL ),
where N S denotes the set of all the substrate nodes, LS denotes the set of all
the substrate links, ASN and ASL stand for the attributes of substrate nodes and
links respectively. In consistency with earlier works[8, 37], in this part we consider
computing capability as node attribute and bandwidth capacity as link attribute. Let
P S denote the set of all the loop-free paths in substrate network. Figure 5.7c shows
an example of a substrate network, where a circle denotes a substrate node, and a line
connecting two circles denotes a substrate link. The number in a square box denotes
the CPU(computing) capacity of that node, and the number next to a substrate link
denotes the bandwidth of that link.
Similarly, we also use an undirected graph GV = (N V , LV , CN V , C V ) to describe
L
V
a virtual network request, where N denotes the set of all the virtual nodes in the
182 5 Intelligent Network Resource Management

request, LV denotes the set of all the virtual links in the request, CN V and C V stand
L
for the constrains of virtual nodes and links respectively. To map a virtual node to
a substrate node, the computing capacity of the substrate node must be higher than
that is required by the virtual node. To map a virtual link to a set of substrate links,
the bandwidth of each substrate link must be higher than that is required by the
virtual link. Figure 5.7a and b show two different virtual requests. Additionally, we
use t to denote the arrival time of a virtual request, and use td to denote the duration
of the virtual request.
When a virtual request arrives, the objective is to find a solution to allocate
different kinds of resources in the substrate network to the request while satisfying
the requirements of the request. If such a solution exists, then the mapping process
will be executed, and the request will be accepted. Otherwise the request will be
rejected or delayed. The virtual network embedding process can be formulated as
a mapping M from GV to GS : GV (N V , LV ) → GS (N , P ), where N ⊂ N S ,
P ⊂ P S.
The main goal of virtual network embedding is to accept as many requests
as possible to achieve maximum revenue for an ISP, when the arrival of virtual
network requests follows an unknown distribution of time and unknown resource
requirements [47]. Consequently, the embedding algorithm must produce efficient
mapping decisions within an acceptable period. As shown in Fig. 5.7, virtual nodes
a and b in request 1 are mapped to substrate nodes E and G respectively, and virtual
nodes c, d and e in request 2 are mapped to substrate nodes A, C and D respectively.
Note that the embedding result of request 1 is not optimal. For example, the cost of
bandwidth in the substrate network can be significantly reduced by moving a to F.
To determine the performance of embedding algorithms, most works use certain
metrics such as a long-term average revenue, a long-term acceptance ratio, and
a long-term revenue to cost ratio. The revenue measures the profit of an ISP for
accepting a certain virtual request, and it depends on the amount of requested
resources and the duration of it. Similar to the earlier works presented in [8, 37],
we define the revenue of accepting a virtual network request as follows:
 
R(Gv , t, td ) = td · [ CP U (nV ) + BW (l V )] (5.38)
nV ∈N V l V ∈N V

where CP U (nV ) and BW (l V ) denote the computing resource that a virtual node
nV requires and the bandwidth resource that a virtual link l V requires respectively.
As shown in the formula, virtual requests having more resources requirements or
lasting longer have more revenue.
The cost function measures the efficiency of utilizing substrate network
resources. We define the cost of accepting a virtual request as follows:
 
C(Gv , t, td ) = td · [ BW (l V )] (5.39)
l V ∈N V l V ∈N V
5.2 Virtual Network Embedding Based on Policy Network and Reinforcement. . . 183

where P (l V ) denotes the set of substrate links where virtual link l V is embedded.
C(GV , t, td ) computes the actual consumption of bandwidth resource for embed-
ding request GV . When accepting a virtual request, the CPU consumption is always
fixed, but the bandwidth consumption may vary depending on the performance of
embedding algorithm discussed above.
Following the works presented in [8, 37, 49], we use a long-term average revenue
to evaluate the overall performance of our embedding method defined as:
T v
t=0 R(G , t, td )
lim (5.40)
T →∞ T

where T is the time elapsed. A higher long-term average revenue leads to a higher
profit for the ISP. Another important metric to evaluate the mapping algorithm is a
long-term acceptance ratio, which means the ratio of accepted requests to the total
number of requests arrived. A higher long-term acceptance ratio means the proposed
algorithm manages to serve more virtual requests.
Finally, a better utilization of substrate network resources would lead to a high
long-term average revenue with comparatively low cost of substrate network. The
long-term revenue to cost ratio, defined as follows, measures the utilization of
substrate network resources:
T
R(Gv , t, td )
lim Tt=0 (5.41)
T →∞ v
t=0 C(G , t, td )

A higher long-term revenue to cost ratio shows that the proposed algorithm is
able to generate more profit with a comparatively less cost to network resources.
We will use these metrics mentioned above to evaluate the performance of our
embedding method in the following subsections.

5.2.2 Embedding Algorithm

In this subsection, we present the details of the proposed policy network based
reinforcement learning algorithm. Specifically, we apply the reinforcement learning
agent in the node mapping stage to derive the probabilities of choosing nodes. The
agent takes a feature matrix extracted from the substrate network as input, and makes
decisions based on a policy network which is trained from historical data.

5.2.2.1 Feature Extraction

Every substrate node has several attributes, such as CPU capacity and the total
amount of bandwidth of the adjacent links [51]. A thorough knowledge of substrate
184 5 Intelligent Network Resource Management

network is crucial for the reinforcement learning agent to establish a basic under-
standing of its state and generate efficient mapping [52, 56]. To facilitate the agent
to choose the substrate nodes, we need to extract features of each substrate node and
use them as input to the policy network.
We extract four features for each substrate node listed as follows:
• Computing capacity (CPU): The CPU capacity of a substrate node nS has a large
impact on its availability. The substrate nodes with a higher computing capacity
are likely to host more virtual nodes.
• Degree (DEG): The degree of a substrate node nS indicates the number of links
connected to it. A substrate node with more adjacent links is more likely to find
paths to other substrate nodes.
• Sum of bandwidth (SU M (BW ) ): Every substrate node is connected to a set of
links. A substrate node nS has a sum of bandwidth resources of its neighboring
links:

SU M (BW ) (nS ) = BW (l S ) (5.42)
l S ∈L(nS )

where L(nS ) is the neighboring links of nS and BW (l S ) is the bandwidth

resource of a substrate link l S . When a substrate node has access to more
bandwidth, mapping a virtual node to it may lead to better link mapping options.
• Average distance to other host nodes AV G(DST ) : When mapping a virtual node,
we also take into consideration the positions where other virtual nodes in the
same request are mapped. By choosing a substrate node close to those already
mapped, the cost of substrate link bandwidth can be reduced. We measure the
distance between two substrate nodes in terms of the number of links along
the shortest path. The shortest path is computed following the Floyd—Warshall
algorithm [53]. We take an average of the distance from a substrate node nS to
another host nodes NS for the same request:

 S
nS ∈N DST (nS ,
nS )
AV G(DST ) (nS ) = (5.43)
S | + 1
|N

nS ) is the distance from node ns to node 

where DST (nS , nS .
In fact, the features that we can extract from the substrate nodes would be
far more than listed above. More features would bring more information about
the substrate network which leads to a better performance of the learning agent.
It should be noted that extracting more features from the substrate network adds
complexity in computation.
After extracting the features of the kth substrate node nSk , we take their
normalized values and concatenate them into a feature vector vk :

vk = (CP U (nSk ), DEG(nSk ), SU M (BW ) (nSk ), AV G(DST ) (nSk ))T (5.44)

5.2 Virtual Network Embedding Based on Policy Network and Reinforcement. . . 185

The purpose of normalization is to accelerate the training process and enable the
agent to converge quickly. We concatenate all feature vectors of substrate nodes to
produce a feature matrix Mf where each row is a feature vector of a certain substrate
node:

Mf = (v1 , v2 · · · v|N S | )T (5.45)

The feature matrix serves as an input to the learning agent. The feature matrix is
updated along with the changing substrate network from time to time.

5.2.2.2 Policy Network

In this work, we implemented an artificial neural network called policy network as

the learning agent. It takes the feature matrix as input and outputs the probabilities
of mapping virtual nodes to substrate nodes.
For simplicity, we build a simple policy network with basic elements of an
artificial neural network as shown in Fig. 5.8. The policy network contains an input
layer, a convolutional layer, a softmax layer and finally a node filter. For each virtual
node that requires a mapping, we use the policy network to choose a substrate node
for it.
At the input layer, we compute the feature matrix and deliver it to the policy net-
work. The policy network then passes the input feature matrix into a convolutional
layer with one convolution kernel, where the policy network evaluates the resources
of each substrate node. The convolutional layer performs a convolution operation
on the input to produce a vector representing the available resources of each node:

hck = ω · vk + b (5.46)

Fig. 5.8 Policy network

186 5 Intelligent Network Resource Management

where hck is the kth output of the convolutional layer, ω is the convolution kernel
weight vector, and b is bias.
Then the vector is transmitted to a softmax layer to produce a probability for each
node which indicates the likelihood of yielding a better result if mapping a virtual
node to it. For the kth node, the probability pk is computed as:
c
e hk
pk =  h c (5.47)
ie
i

The softmax function is a generalization of the logistic regression. It turns the

n-dimensional vector into real values between 0 and 1 that add up to 1. The output
of the softmax function indicates a probability distribution over n different possible
mappings. Some of the nodes are not able to host the virtual node in concern because
they do not have enough computing resources. We add a filter to choose a set of
candidate nodes with enough CPU capacities.

5.2.2.3 Training and Testing

We first randomly initialize the parameters in the policy network, and train it for
several epochs. For every virtual node in each iteration, a feature matrix is extracted
from the substrate network which serve as input to the policy network. The policy
network outputs a set of available substrate nodes as well as a probability for
each node. The probability of each node represents the likelihood that mapping a
virtual node to it will yield a better result. In the training stage, we cannot simply
select the node with a maximal probability as the host because that the model is
randomly initialized, which means the output could be biased and better solutions
might exist. In other words, we need to reach a balance between the exploration
of better solutions and the exploitation of current model. To this end, we generate
a sample from the set of available substrate nodes according to their probability
distribution that the policy network outputs, and select a node as the host. We repeat
this process until all the virtual nodes in a virtual request are assigned and proceed
to link mapping. If no substrate node is available, the mapping fails due to a lack
of resources. For link mapping, we apply a breadth-first search to find the shortest
paths between each pair of nodes.
In supervised learning, each piece of data in the training set corresponds to a
label indicating the desired output of the model. With each output from model and
the corresponding label, a loss value is computed which measures the deviation
between them. The loss value for each piece of data in the training set sums up to an
aggregated loss value, and the training stage aims to minimize the aggregated loss
value. However, in reinforcement learning tasks such as virtual network embedding,
data in the training set does not have corresponding labels. The learning agent relies
on reward signals to know if it is working properly. A big reward signal informs the
learning agent that its current action is effective and should be continued. A small
5.2 Virtual Network Embedding Based on Policy Network and Reinforcement. . . 187

reward signal or even a negative reward signal shows that the current action is
erroneous and should be adjusted. The choice of reward is critical in reinforcement
learning as it directly influences the training process and determines the final policy.
Here, we use the revenue to cost ratio of a single virtual request as the reward
for every virtual node in this request because this metric represents the utilization
efficiency of the substrate resources. Then we apply policy gradient method to train
the policy network.
The actual implementation of the proposed algorithm is non-trivial since we
cannot provide each output with a label. As a result, we temporarily consider every
decision that the agent makes to be correct by introducing a hand-crafted label into
our policy network. Assume that we choose the ith node, then the hand-crafted label
in policy network would be a vector y filled with zeros except the ith position which
is one. Then we calculate the cross-entropy loss:

L(y, p) = − yi log(pi ) (5.48)
i

where yi and pi are the ith element of hand-crafted label and the output of
policy network respectively. We use backpropagation to compute the gradients
of parameters in the policy network. Since we use hand-crafted label, we stack
the gradients gf rather than applying them immediately. If our algorithm fails to
embed a virtual request, the corresponding stacked gradients will be aborted since
we cannot determine the reward signal. If a virtual request has been successfully
mapped, we compute its revenue to cost ratio as a reward r. Then we multiply the
stacked gradients by using the reward and an adjustable learning rate α to achieve
the final gradients:

g = α · r · gf (5.49)

The learning rate α is introduced to control the magnitude of gradients and the
computation speed of training. If the gradients are too large, the model becomes
unstable and may not improve through the training process. On the other hand, too
small gradients make training extremely slow. Therefore the learning rate needs
to be tuned carefully. It can be observed from Eq. (5.49) larger rewards make the
corresponding gradients more significant than small ones. As a result, the choices
that lead to larger rewards have larger impact on the learning agent, making it more
prone to make similar decisions. When we stack a batch of gradients, we apply
them to parameters and update the policy network. There are two reasons for batch
updating—one is that parameter updating normally takes a long time, but doing that
in batches speeds up this process. Another reason is that batch updating averages
over the gradients and is more stable. The training process is shown in Algorithm 4.
Lines 7–10 show node mapping stage where we compute the gradients in line 10,
lines 11–13 show the link mapping stage.
188 5 Intelligent Network Resource Management

Algorithm 4 Training process

Input: Number of epochs, numEpoch; Learning rate, α; Training set;
Output: Trained parameters in policy network;
1: Initialize all the parameters in policy network;
2: while iteration < numEpoch do
3: for req ∈ trainingSet do
4: counter=0;
5: for node ∈ req do
6: Mf =getFeatureMatrix();
7: p=policyNet.getOutput(Mf ) //Get the probability distribution from policy network;
8: host=sample(p) //Sample from the probability distribution to choose a node as host;
9: computeGradient(host);
10: end for
11: if isMapped(∀ node ∈ req) then
12: bfsLinkMap(req);
13: end if
14: if isMapped(∀ node ∈ req, ∀ link ∈ req) then
15: reward=revToCost(req) //Compute revenue to cost ratio;
16: multiplyGradient(reward, α) //Compute the final gradients;
17: else
18: clear the stacked gradients;
19: end if
20: ++counter;
21: if counter reach the batch size then
22: apply gradients to parameters;
23: counter=0;
24: end if
25: end for
26: ++iteration;
27: end while

In the testing stage, we apply a greedy strategy where we directly choose the
node with the highest probability as the host. The testing algorithm is shown in
Algorithm 5.

Algorithm 5 Testing process

Input: testing set;
Output: long-term average revenue, acceptance ratio, long-term revenue to cost ratio;
1: Initialize all the parameters in policy network;
2: for req ∈ testSet do
3: for node ∈ req do
4: Mf =getFeatureMatrix();
5: host=maxProbablity(p) //Greedy strategy;
6: end for
7: bfsLinkMap(req);
8: if isMapped(∀ node ∈ req, ∀ link ∈ req) then
9: signal(SUCCESS);
10: end if
11: end for
5.2 Virtual Network Embedding Based on Policy Network and Reinforcement. . . 189

5.2.3 Evaluation

We conducted a number of simulation tests to evaluate the performance of the

proposed reinforcement learning algorithm and compared with other embedding
algorithms.
We employed GT-ITM tool [54] to generate a substrate network with 100 nodes
and approximately 550 links, which is about a middle-sized ISP. The computing
capacity of every substrate node is a real number that follows a uniform distribution
between 50 and 100, and the bandwidth of every link is a real number that follows
a uniform distribution between 20 and 50, which is similar to parameters presented
in related work [8, 37].
We also generated a number of virtual requests, each with 2 to 10 virtual nodes.
The computing capacity requirement of every virtual node followed a uniform
distribution between 0 and 50 units. The bandwidth requirement of every virtual
link follows a uniform distribution between 0 and 50 units. Virtual nodes were
connected with a probability of 0.5 forming an average of n(n−1) 4 virtual links,
where n is the number of nodes. The virtual requests arrived following a Poisson
distribution with an average of 4 requests over 100 time units. The duration of every
requests followed an exponential distribution with an average of 1000 time units.
We generate a timeline that lasted around 50,000 time units containing about 2000
requests. We divided the requests equally into two sets—training set and testing set.
We employed TensorFlow [55] to build the policy network. We first assembled
the layers mentioned in Sect. 5.2.2 and followed a normal distribution to initialize
their parameters. Then we defined a tensor for gradient of every training step using
the compute_gradients method of stochastic gradient descent [57] optimizer.
When doing batch updates, we added all the gradients for each parameter and
multiply them by a reward, and called the apply_gradients method of SGD
optimizer to update all parameters in policy network. We train our agent for 100
epochs using gradient descent with a learning rate of 0.005. The batch size was set
to 100 which means the parameters are updated every 100 requests.

5.2.3.1 Effectiveness of Reinforcement Learning

Compared to supervised learning, reinforcement learning has proven to be hard to

train. It may take a very long time for the learning agent to stabilize, especially for a
complicated problem such as virtual network embedding. We run the learning agent
on the training data set for 100 epochs and observe its performance. Figure 5.9a
shows the change of a long-term average revenue, the acceptance ratio and a long-
term revenue to cost ratio during the training process. In the very beginning, the
learning agent performs poorly because all the parameters in the policy network are
initialized randomly. As the training goes, the random sampling allows the learning
190 5 Intelligent Network Resource Management

Fig. 5.9 Training process. (a) Performance on training set. (b) Loss on training set

agent to explore different possibilities. The learning agent may find a good solution
occasionally and receive a great reward which helps the policy network to learn to
make better decisions. Consequently, the performance starts to get better, proving
the effectiveness of reinforcement learning on the task. The exploration strategy
sometimes leads our agent into bad choices causing a fluctuation in its performance
as the training proceeds. But such cases will lead to small rewards and have small
5.2 Virtual Network Embedding Based on Policy Network and Reinforcement. . . 191

impact on the learning agent. In the later stage of training process, the performance
stopped improving due to the limited capacity of functional complexity that the
policy network can handle. Eventually the learning agent reaches a certain point,
and the performance stabilizes in a range. Figure 5.9b shows the cross-entropy loss
during the training process. Clearly, the loss decreases through the training stage
and eventually starts to stabilize in the last 10 epochs.
The result shows that the proposed reinforcement learning based algorithm is
getting better performance as the training goes, which means the learning agent can
adapt itself to training data.
We have proved that the reinforcement learning method can improve the
performance of the learning agent on training set. But it is still unclear if the learning
agent actually learns how to optimize node mapping, or simply adjusts to existing
data. In order to test the generalization ability of the learning agent, we separated a
testing data set that consisted of different requests from the training set and run the
learning agent on it.
Different from the training process, we run the learning agent without a random
sampling and applied a greedy strategy to choose a node with the maximal proba-
bility. The performance over time on the testing data set is shown in Fig. 5.10. We
compare the learning agent with another two rule-based node ranking algorithms.
The first one is a baseline algorithm proposed in [37] using equation:

H (nS ) = CP U (nS ) BW (LS ) (5.50)
l S ∈L(nS )

to rank substrate nodes, where H (ns ) measures the availability of substrate node ns .
The other is proposed in [8] using NodeRank algorithm to measure the importance
of nodes. All the three methods followed the same breadth-first search link mapping
algorithm. We measured the performance of these methods with three metrics
mentioned in Sect. 5.2.1—a long-term average revenue, an acceptance ratio and a
long-term revenue to cost ratio.
At the beginning, the performance of all the three algorithms on a long-term
average revenue and an acceptance ratio decrease because the amount of resources
of the substrate network decreases as more requests arrive. The long-term revenue
to cost ratio is stable because it is irrelevant to the amount of available resources.
Then the performance of all algorithms on all metrics starts to stabilize because the
resources of the substrate network is depleted. The results in Fig. 5.10 show that
the learning agent outperforms the other two algorithms in all the three metrics.
192 5 Intelligent Network Resource Management

Fig. 5.10 Performance on testing set

The training data set and testing data set consist of different requests, but the
learning agent is able to perform well on both data sets. The conclusion is that the
learning agent does not simply adjust itself to the training set, rather it is actually
capable of generalizing from the training process to acquire knowledge about the
substrate network and node mapping. Note that the performance improved evidently
compared to the result on the training data set, because the exploration in training
process may lead to bad embedding results.

5.2.3.2 Stress Tests

We further expanded our experiments by increasing resource demands to examine

the performance of the proposed algorithm under more stress. By doing this we also
expected to find the circumstances where our algorithm is robust.
Figure 5.11a shows the performances of the three algorithms when we doubled
the CPU requirements of the virtual nodes. Figure 5.11b shows the results when
the bandwidth of virtual links were uniformly distributed between 25 and 50
units. As shown in Fig. 5.11, the proposed reinforcement learning based algorithm
performs better in the later circumstance, and outperforms the other two methods
against the three metrics. In the computing-intensive environment however, the
5.2 Virtual Network Embedding Based on Policy Network and Reinforcement. . . 193

Fig. 5.11 Stress tests. (a) Performance in a computing-intensive environment. (b) Performance in
a bandwidth-intensive environment
194 5 Intelligent Network Resource Management

proposed algorithm achieves similar performance to the other two methods in terms
of a long-term revenue to cost ratio while getting better results in the long-term
average revenue and acceptance ratio. The proposed reinforcement learning based
algorithm works in node mapping phase, but larger CPU requirements means less
nodes with enough computing resource to choose from, which leads to relatively
worse performance in computing-intensive environment. The conclusion is that the
proposed algorithm can achieve comparatively better performance for bandwidth-
intensive requests rather than computing-intensive ones.

5.3 Summary

In this chapter, we discuss the main challenge of intelligent network resource

management and introduced several reinforcement learning algorithms. we first
propose a reinforcement learning based dynamic attribute matrix representation
(RDAM) algorithm for virtual network embedding. Then, we design and implement
a policy network based on reinforcement learning to make node mapping decisions.

References

1. J. L. Chen, Y. W. Ma, H. Y. Kuo, and C. S. Yang, “Software-defined network virtualization

platform for enterprise network resource management,” IEEE Transactions on Emerging
Topics in Computing, vol. 4, no. 2, pp. 179–186, 2016.
2. J. Lu and J. Turner, “Efficient mapping of virtual networks onto a shared substrate,” Washington
University in St Louis, 2006.
3. Z. Zhen, T. Jiang, W. Zhang, H. Yao, and S. Xiao, “Analyzing speech of patients with vocal
polyps based on channel parameters and fuzzy logic systems,” Computers & Mathematics with
Applications, vol. 62, no. 7, pp. 2834–2842, 2011.
4. D. G. Andersen, “Theoretical approaches to node assignment,” Computer Science Department,
2002.
5. Zhao, Chenglin, Haipeng, Zhou, and Zheng, “Optimization of multiband spectrum sensing for
cognitive radio networks under sensing capability constrains,” China Communications, vol. 7,
no. 5, pp. 129–136, 2010.
6. C. Jiang, C. Yan, K. J. R. Liu, and R. Yong, “Network economics in cognitive networks,” IEEE
Communications Magazine, vol. 53, no. 5, pp. 75–81, 2015.
7. H. Li, M. Dong, K. Ota, and M. Guo, “Pricing and repurchasing for big data processing in
multi-clouds,” IEEE Transactions on Emerging Topics in Computing, vol. 4, no. 2, pp. 266–
277, 2016.
8. X. Cheng, S. Su, Z. Zhang, H. Wang, F. Yang, Y. Luo, and J. Wang, “Virtual network
embedding through topology-aware node ranking,” Acm Sigcomm Computer Communication
Review, vol. 41, no. 2, pp. 38–47, 2011.
9. Z. Tong, C. Yan, C. Jiang, and K. J. R. Liu, “Pricing game for time mute in femto-macro
co-existent networks,” IEEE Transactions on Wireless Communications, vol. 14, no. 4, pp.
2118–2130, 2015.
10. C. Jiang, X. Wang, W. Jian, H. H. Chen, and R. Yong, “Security in space information networks,”
IEEE Communications Magazine, vol. 53, no. 8, pp. 82–88, 2015.
References 195

11. C. Jiang, C. Yan, and K. J. R. Liu, “Data-driven optimal throughput analysis for route selection
in cognitive vehicular networks,” IEEE Journal on Selected Areas in Communications, vol. 32,
no. 11, pp. 2149–2162, 2014.
12. C. Jiang, C. Yan, G. Yang, and K. J. R. Liu, “Indian buffet game with negative network exter-
nality and non-bayesian social learning,” IEEE Transactions on Systems Man & Cybernetics
Systems, vol. 45, no. 4, pp. 609–623, 2013.
13. C. Jiang, Y. Chen, Q. Wang, and K. J. R. Liu, “Data-driven auction mechanism design in iaaS
cloud computing,” IEEE Transactions on Services Computing, vol. 11, no. 5, pp. 743–756,
2018.
14. C. Jiang, Y. Chen, Y. H. Yang, C. Y. Wang, and K. J. R. Liu, “Dynamic chinese restaurant
game: Theory and application to cognitive radio networks,” IEEE Transactions on Wireless
Communications, vol. 13, no. 4, pp. 1960–1973, 2014.
15. C. Jiang, Y. Chen, Y. Gao, and K. J. R. Liu, “Indian buffet game with negative network exter-
nality and non-bayesian social learning,” IEEE Transactions on Systems Man & Cybernetics
Systems, vol. 45, no. 4, pp. 609–623, 2015.
16. L. Feng, C. Jiang, J. Du, Y. Jian, R. Yong, Y. Shui, and M. Guizani, “A distributed gateway
selection algorithm for uav networks,” IEEE Transactions on Emerging Topics in Computing,
vol. 3, no. 1, pp. 22–33, 2017.
17. Y. Kawamoto, H. Takagi, H. Nishiyama, and N. Kato, “Efficient resource allocation utilizing
q-learning in multiple ua communications,” IEEE Transactions on Network Science &
Engineering, vol. PP, no. 99, pp. 1–1.
18. X. Lei, C. Jiang, C. Yan, R. Yong, and K. J. R. Liu, “Privacy or utility in data collection?
a contract theoretic approach,” IEEE Journal of Selected Topics in Signal Processing, vol. 9,
no. 7, pp. 1256–1269, 2015.
19. L. Zhang, H. Yao, H. Liu, and Z. Zhou, “A novel ultra-wide band signal generation
scheme based on carrier interference and dynamics suppression,” Eurasip Journal on Wireless
Communications & Networking, vol. 2010, no. 1, p. 10, 2010.
20. H. Yao, X. Chen, M. Li, P. Zhang, and L. Wang, “A novel reinforcement learning algorithm for
virtual network embedding,” Neurocomputing, vol. 284, pp. 1–9, 2018.
21. M. Belkin and P. Niyogi, “Laplacian eigenmaps and spectral techniques for embedding and
clustering,” Advances in Neural Information Processing Systems, vol. 14, no. 6, pp. 585–591,
2009.
22. U. V. Luxburg, “A tutorial on spectral clustering,” Statistics & Computing, vol. 17, no. 4, pp.
395–416, 2007.
23. G. Peters and J. H. Wilkinson, “ax = bx and the generalized eigenproblem,” Siam Journal on
Numerical Analysis, vol. 7, no. 4, pp. 479–492, 1970.
24. G. W. Stewart and J. G. Sun, “Matrix perturbation theory,” 1990.
25. A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classification with deep convolutional
neural networks,” in International Conference on Neural Information Processing Systems,
2012, pp. 1097–1105.
26. C. Jiang, C. Yan, R. Yong, and K. J. R. Liu, “Maximizing network capacity with optimal source
selection: A network science perspective,” IEEE Signal Processing Letters, vol. 22, no. 7, pp.
938–942, 2014.
27. K. Arulkumaran, M. P. Deisenroth, M. Brundage, and A. A. Bharath, “A brief survey of deep
reinforcement learning,” 2017.
28. D. Drutskoy, E. Keller, J. Rexford, Scalable network virtualization in software-defined
networks, IEEE Internet Computing 17 (2) (2013) 20–27.
29. N. Zhang and H. P. Yao, “Overview of euicc remote management technology,” Telecom
Engineering Technics & Standardization, 2012.
30. R. Jain, S. Paul, Network virtualization and software defined networking for cloud computing:
a survey, Communications Magazine IEEE 51 (11) (2013) 24–31.
31. A. Fischer, J. F. Botero, M. T. Beck, H. D. Meer, X. Hesselbach, Virtual network embedding:
A survey, IEEE Communications Surveys & Tutorials 15 (4) (2013) 1888–1906.
196 5 Intelligent Network Resource Management

32. C. Liang, F. R. Yu, Wireless network virtualization: A survey, some research issues and
challenges, IEEE Communications Surveys & Tutorials 17 (1) (2015) 358–380.
33. N. M. K. Chowdhury, R. Boutaba, Network virtualization: state of the art and research
challenges, IEEE Communications magazine 47 (7) (2009) 20–26.
34. H. Zhang, C. Jiang, J. Cheng, and V. C. M. Leung, “Cooperative interference mitigation and
handover management for heterogeneous cloud small cell networks,” Wireless Communica-
tions IEEE, vol. 22, no. 3, pp. 92–99, 2015.
35. N. M. M. K. Chowdhury, R. Boutaba, A survey of network virtualization, Computer Networks
54 (5) (2010) 862–876.
36. Y. Zhu, M. Ammar, Algorithms for assigning substrate network resources to virtual network
components, in: INFOCOM 2006. IEEE International Conference on Computer Communica-
tions. Proceedings, 2007, pp. 1–12.
37. M. Yu, Y. Yi, J. Rexford, M. Chiang, Rethinking virtual network embedding: substrate support
for path splitting and migration, Acm Sigcomm Computer Communication Review 38 (2)
(2008) 17–29.
38. L. Xu, C. Jiang, J. Wang, J. Yuan, and Y. Ren, “Information security in big data: Privacy and
data mining,” IEEE Access, vol. 2, no. 2, pp. 1149–1176, 2017.
39. N. M. M. K. Chowdhury, M. R. Rahman, R. Boutaba, Virtual network embedding with
coordinated node and link mapping, Proceedings - IEEE INFOCOM 20 (1) (2009) 783–791.
40. C. Jiang, N. C. Beaulieu, Z. Lin, R. Yong, M. Peng, and H. H. Chen, “Cognitive radio networks
with asynchronous spectrum sensing and access,” Network IEEE, vol. 29, no. 3, pp. 88–95,
2015.
41. M. A. Ying-Jie, Z. Zhou, H. E. Wen-Cai, J. Zhang, and H. P. Yao, “Cognitive uwb orthogonal
pulses design and its performance analysis,” Transactions of Beijing Institute of Technology,
vol. 31, no. 5, pp. 583–588, 2011.
42. S. Shanbhag, A. R. Kandoor, C. Wang, R. Mettu, T. Wolf, Vhub: Single-stage virtual network
mapping through hub location, Computer Networks 77 (2015) 169–180.
43. D. Silver, A. Huang, C. J. Maddison, A. Guez, L. Sifre, G. Van Den Driessche, J. Schrittwieser,
I. Antonoglou, V. Panneershelvam, M. Lanctot, et al., Mastering the game of go with deep
neural networks and tree search, Nature 529 (7587) (2016) 484–489.
44. C. Jiang, C. Yan, K. J. R. Liu, and R. Yong, “Optimal pricing strategy for operators in cognitive
femtocell networks,” Wireless Communications IEEE Transactions on, vol. 13, no. 9, pp. 5288–
5301, 2014.
45. V. Mnih, K. Kavukcuoglu, D. Silver, A. A. Rusu, J. Veness, M. G. Bellemare, A. Graves,
M. Riedmiller, A. K. Fidjeland, G. Ostrovski, Human-level control through deep reinforcement
learning, Nature 518 (7540) (2015) 529.
46. S. Mozer, M. C, M. Hasselmo, Reinforcement learning: An introduction, Machine Learning
8 (3-4) (1992) 225–227.
47. L. Meng, F. R. Yu, P. Si, H. Yao, E. Sun, and Y. Zhang, “Energy-efficient m2m communi-
cations with mobile edge computing in virtualized cellular networks,” in IEEE International
Conference on Communications, 2017.
48. C. Jiang, C. Yan, and K. J. R. Liu, “Evolutionary dynamics of information diffusion over social
networks,” IEEE Transactions on Signal Processing, vol. 62, no. 17, pp. 4573–4586, 2014.
49. X. Jin, P. Zhang, and H. Yao, “A communication framework between backbone satellites and
ground stations,” in International Symposium on Communications & Information Technolo-
gies, 2016.
50. S. Haeri, L. Trajkovic, Virtual network embedding via monte carlo tree search., IEEE
Transactions on Cybernetics (99) (2017) 1–12.
51. R. Mijumbi, J. L. Gorricho, J. Serrat, M. Claeys, F. D. Turck, S. Latre, Design and evaluation
of learning algorithms for dynamic resource management in virtual networks, in: Network
Operations and Management Symposium, 2014, pp. 1–9.
52. H. Zhang, C. Jiang, N. C. Beaulieu, X. Chu, X. Wen, and M. Tao, “Resource allocation
in spectrum-sharing ofdma femtocells with heterogeneous services,” IEEE Transactions on
Communications, vol. 62, no. 7, pp. 2366–2377, 2014.
References 197

53. S. Hougardy, The Floyd–Warshall algorithm on graphs with negative cycles, Information
Processing Letters 110 (8) (2010) 279–281.
54. M. Thomas, E. W. Zegura, Generation and analysis of random graphs to model internetworks,
College of Computing volume 63 (4) (1994) 413–442(30).
55. M. Abadi, A. Agarwal, P. Barham, E. Brevdo, Z. Chen, C. Citro, G. S. Corrado, A. Davis,
J. Dean, M. Devin, et al., Tensorflow: Large-scale machine learning on heterogeneous
distributed systems, arXiv preprint arXiv:1603.04467.
56. Q. Chao, C. Zhao, H. Yao, F. Xu, and F. R. Yu, “Why did you opt to switch off me? big data
for green software defined networking,” in Globecom Workshops, 2017.
57. L. Bottou, Online algorithms and stochastic approximations, in: D. Saad (Ed.), Online Learning
and Neural Networks, Cambridge University Press, Cambridge, UK, 1998, revised, October
2012.
Chapter 6
Intention Based Networking
Management

Compared to traditional networking which using command-line interfaces, intent-

based networking abstracts network complexity and improves automation by
eliminating manual configurations. It allows a user or administrator to send a
simple request—using natural language—to plan, design and implement/operate
the physical network which can improve network availability and agility. For
example, an IT administrator can request improved voice quality for its voice-
over-IP application, and the network can respond. For intent-based networking,
the translation and validation system take a higher-level business policy (what) as
input from end users and converts it to the necessary network configuration (how)
by natural language understanding technology. In this chapter, we focus on how
artificial intelligence technology can be used in the natural language understanding
in translation and validation system. We firstly propose an effective model for
the similarity metrics of English sentences. In the model, we first make use of
word embedding and convolutional neural network (CNN) to produce a sentence
vector and then leverage the information of the sentence vector pair to calculate
the score of sentence similarity. Then, we propose the SM-CHI feature selection
method based on the common method used in Chinese text classification. Besides,
the improved CHI formula and synonym merging are used to select feature words
so that the accuracy of classification can be improved and the feature dimension
can be reduced. Finally, we present a novel approach which considers both the
semantic and statistical information to improve the accuracy of text classification.
The proposed approach computes semantic information based on HowNet and
statistical information based on a kernel function with class-based weighting.

© Springer Nature Switzerland AG 2019 199

H. Yao et al., Developing Networks using Artificial Intelligence,
Wireless Networks, https://doi.org/10.1007/978-3-030-15028-0_6
200 6 Intention Based Networking Management

6.1 CNN Based Sentence Similarity Model

Sentence similarity measurement is a fundamental problem in the research of natural

language processing (NLP) and has been used in many language processing tasks
such as query ranking, question answering, paraphrase identification and paraphrase
generation, to cite a few [57]. Understanding the similarity between sentences is
not a simple task for machines, because both semantic information and syntactic
information should be taken into consideration. The distributed word embedding
has been explicitly encoded many linguistic regularities and patterns in the works of
Mikolov et al. [1–3] and others, and it has been widely utilized to extract semantic
features of sentences. The syntax parsing is a challenge task due to its complexity,
therefore it is a good idea that makes machines hidden learn grammar rules from
corpus. Recently, as the increasing interest in neural networks, the idea is realized
by means of neural network-based models.
In 2008, Collobert and Weston [5] proposed a convolutional neural network-
based architecture for modeling sentences and used it in several learning tasks
such as named entity tags, semantic roles, semantically similar words and so on.
In their studies, words were embedded into the multi-dimension space. Inspired by
their works, other researchers utilized word embedding and convolutional neural
network in paraphrase identification [6, 7], sentences classification [4, 10, 12, 13]
and semantic similarity measurement [8, 10, 14, 16]. Recurrent neural network,
a time sequence neural network with variable-length input, is good choice for
sequence modeling tasks like modeling sentences. Socher et al. [19] applied RNN
in their task that finding and describing images with sentences. The disadvantage of
RNN is its gradients vanishing over long sequence. To avoid the vanishing gradient
problem, the long short-term memory networks (LSTM) were proposed, which is
better in learning long range dependencies. Tai et al. [10] and Mueller et al. [20]
used LSTM for sentence similarity measurement.
However, both long short-term memory networks and deep convolutional neural
networks have a disadvantage of time-consuming when they are utilized for seman-
tic similarity measurement [11]. In addition, deep convolutional neural networks
need to be modified for measuring sentence similarity. The structures of them
universally are complicated. Since the measurement of sentence similarity is a
fundamental task in NLP and more commonly serves as a subtask that aids in
solving other larger tasks, these models that have complicated structure and cost
long time for training will consume extra resources that assigned for other subtasks
and slow down the proceeding of the whole task [15].
In this chapter, we propose a novel model with word embedding and convolu-
tional neural network for sentence similarity measurement. The major contributions
of this part are as follows:
1. To reduce the training time and simplify architecture structure, we propose a
novel sentence similarity model that is different from traditional long short-
term memory networks and deep convolutional neural networks. The main
body of our model is a shallow convolutional neural network without requiring
6.1 CNN Based Sentence Similarity Model 201

fully connected layer meanwhile modifying its structure targeting at the task of
sentence similarity measurement.
2. The sentence with variable length is embedded into a multiple dimensional
space. We extend the method of word embedding from word-level to sentence-
level, with the aim of putting word similarity computation methods into sentence
similarity measurement. Additionally, our model can handle any length sentences
without needing clipping and padding operations.
3. We evaluate our model using different evaluation metrics with two different kinds
of tasks, namely, semantic relatedness task (SemEval 2014, Task 1) and the
Microsoft research paraphrase identification task. The obtained results achieve
a good performance on both tasks.

6.1.1 The Proposed Model WSV-SCNN-SV

In this part, we propose our neural network-based model namely “the word
set vectors-the shallow convolutional neural network-the sentence vector” (WSV-
SCNN-SV). The main body of the model is the shallow convolutional neural
network which takes the group of “word set vectors” as the input feature map and
outputs the sentence representation—the sentence vector. The “word set vector” is
a new idea considering as the improvement of the word embedding. Compared with
word embedding, it contains more sematic information of sentences. We compute
the similarity based on the sentence vectors that learned by the convolutional neural
network instead of employing the convolutional neural network [17].
The organization of this subsection is summarized as follows: First, we introduce
the whole architecture of our model, then we elaborate on each part of our model in
detail.

6.1.1.1 The Overall Architecture of the Model

The hybrid architecture of WSV-SCNN-SV is shown in Fig. 6.1. In the architecture,

there are four parts:
Part a: Sentences Pre-processing Since the raw sentence is not accepted by
machines, it should be processed and encoded before it input into machines. The
pre-process of the sentences includes word segmentation, removal of stop words,
stemming and son on. In our work, we use word embedding to encode sentence in
the way that words from each sentence are represented by their corresponding word
vector after the pre-processing.
Part b: The Input Model We propose a novel method that transforms the sequence
of word vectors into a three-dimensional tensor. The tensor which contains rich
semantic and syntactic information of the sentence is the input feature map of the
next convolutional neural network.
202 6 Intention Based Networking Management

Fig. 6.1 The architecture of the model

Part c: Our Shallow Convolutional Neural Network The work of our convolu-
tional neural network is learning the semantic and syntactic features from the input
tensor and producing the sentence representation. We make some modification to
the convolutional neural network. First, we remove fully connected layer from our
model. After flowing over several convolutional and pooling layers, the tensor is
converted into a vector which extracts rich features from the input tensor. Second,
we apply both k-max pooling and max pooling operation in the model. Third, we
make our convolutional neural network support the input feature map with unfixed
size by means of k-max pooling operation.
Part d: Similarity Computation The sentence vector pair is used to compute the
score of similarity. Many methods are available for similarity measurement between
two vectors. In our work, we use cosine distance, Euclidean distance and Manhattan
distance, respectively.
In summary, the aim of the first three parts of the model is to accomplish the task
that embeds sentences into a high dimensional space. The goal of the last part of our
model is to calculate the score of similarity with the sentence vector pair.

6.1.1.2 Sentences Pre-processing

Since each word is an atomic semantic unit for a sentence, it is of great significance
for machines to capture fine-grained features from sentences [61]. Different sen-
tences may use different words to convey the high similar information, for instance,
“He studies computer science in college” and “His major is CS”. Many remarkable
researches have been done in modeling words like distributed word embedding
which explicitly encode many linguistic regularities and word similarities.
6.1 CNN Based Sentence Similarity Model 203

One of notable works on distributed word embedding is word2vec model

[1, 3]. In word2vec model, the result of a vector calculation vector(“Madrid”) −
vector(“Spain”) + vector(“France”) is closer to vector(“Paris”) than to any other
word vector. It is universal in related papers that base on word2vec to model
sentence measuring similarity. Similarly, in this part, we apply pre-trained word
embedding to model words.
There are many simple methods to produce the sentence representation using
word embedding. Assume wi = (wi1 , wi2 , . . . , wid ) is a d-dimensional word vector
corresponding to the i-th word in the sentence. A sentence contained n words can
be represented as follows:

1
n
S= wi (6.1)
n
i=1

S = (wT1 , wT2 , . . . , wTn ) (6.2)

In prior studies, these sentence representations are input into neural network to finish
final task. Formula (6.1) uses the average value of these word vectors to express the
sentence [23, 25–27, 30, 31], despite this process is very simple, its performance
in practice is not poor. Formula (4.31) concatenates word vectors to produce a two-
dimensional matrix to express the sentence, which is a common method to represent
sentence. Different from formula (4.30), the method of formula (4.31) keeps order
information of words. Kim [13] used the method, and Hu et al. [6] made some
improvement of formula (6.2).
However, our idea is different from the previous work. Our motivation is that
sentence representation should learn from neural network as the word embedding
does. After words in the sentence are represented by their corresponding word
vector, we construct a three dimensional tensor and then put the tensor into the
convolutional neural network to learn the sentence representation.

6.1.1.3 The Input Model Based on “Binary-Gram Word Set Vector”

The word is not the only factor to be taken into consideration in the sentence
representation. The dependencies between word and phrase, phrase and phrase
cannot be ignored. The representation of the sentence in formula (6.2) can be treated
as a feature map with one width, n height and d channels. However, the disadvantage
of the representation is that long-range semantic dependencies between words have
been ignored.
To consider the long-range semantic dependencies in our model, we propose
a new vector dubbed “word set vector”. The idea of word set vector is inspired
by the idea that sentences can be represented by the average of word vectors. We
extend this idea from sentences to phrases and collections of words that are semantic
dependencies with each other. Word set vector w composed with j word vectors
204 6 Intention Based Networking Management

wl1 , wl2 , . . . , wlj is defined as follows:


j
w= λi wli (6.3)
i=1

where the sequence {l1 , l2 , . . . , lj } denotes a subsequence of the sequence

j
{1, 2, . . . , n}, λ denotes the weight of word vector ( i=1 λi = 1, λi > 0).
The word set vector composed by word vectors wl1 , wl2 , . . . , wlj is order
sensitive. The word set vector w composed with j word vectors (allowed repeated
word vectors) is defined as j-gram word set vector. We deem that the parameter λ
should be trained by machines, it can be calculated as follows:

exp(αi )
λi =  , αj ∈ R (6.4)
j exp(αj )

The word set vector uses the weight λ to punish disorder word vectors. A
sentence with n words has nj j-gram word set vectors. A sentence with n words
has n2 d-dimensional binary-gram word sets vectors which can be transformed into
a feature map with n width, n height and d channels. The binary-gram word set
vector is calculated by:


wl1 l2 = λ1 wl1 + λ2 wl2 (6.5)

where l1 , l2 ∈ {1, 2, . . . , n}, λ1 + λ2 = 1, λ1 , λ2 > 0. If l1 = l2 , 

wl1 l2 denotes a
binary-gram word set vector. The weight λ1 , λ2 are calculated by gradient descent.
Algorithm 1 shows how the input feature map is composed. The units of the map
are these binary-gram word set vectors in the sentence.

Algorithm 1 The computation process of the input feature map

Input: Word embeddings, w1 , w2 , . . . , wn
Output: The input feature map, map
1: Initialize the input feature map: map ∈ Rn×n×d to all zero
2: Initialize variables: α1 , α2 to 0.4,0.6
3: λ1 = exp(α1 )/(exp(α1 ) + exp(α2 ))
4: λ2 = exp(α2 )/(exp(α1 ) + exp(α2 ))
5: for i = 1 to n do
6: for j = 1 to n do
7: for depth = 1 to d do
8: map[i][j ][depth] = λ1 · wi [depth] + λ2 · wj [depth]
9: end for
10: end for
11: end for
6.1 CNN Based Sentence Similarity Model 205

Fig. 6.2 The input feature

map of the convolutional
neural network where
ij
w k = λ w k + λ w k denotes
1 i 2 j
the k-th dimensional value of
word set vector that is
composed of word vector wi
and wj

Figure 6.2 illustrates the feature map that is composed of n2 word set vectors. It
is one of creative works in the part that extending the feature map of sentences from
a two-dimensional matrix to a three-dimensional tensor. As the feature map is the
same as the image data in form, we hope that the next convolutional neural network
can fully extract sentence features from the input feature map as it does on images.

6.1.1.4 Our Shallow Convolutional Neural Network

The complete multi-layer’s convolutional neural network should contain fully

connected layer just like Fig. 6.3a. In our convolutional neural network, we remove
the fully connected layer from the network as it is not essential for learning sentence
vectors and consumes time for training. Our proposed convolutional neural network
only contains layers that consist of convolutional sublayers and pooling sublayers.
In addition, a sum function that computes the sentence vectors is applied at the
output of the last layer.
Convolutional Sublayers The formula of convolution is as follows:

Xjl = φ( Xil−1 ∗ Kijl + bjl ) (6.6)
i∈Mj

where Xjl denotes the j -th map on the l-th sublayer, φ(x) denotes the active
function, Mj denotes the number of maps on the (l − 1)-th sublayer, K l denotes
the filter of the l-th sublayer, bjl denotes the bias.
We do not make any modification to convolutional sublayers. As can be seen in
Fig. 6.3a, zero padding (dashed blocks are zero padding) is used in the convolutional
206 6 Intention Based Networking Management

Fig. 6.3 The convolutional neural network, k-max pooling. (a) The convolutional neural network
with max-pooling operation. (b) The convolutional neural network with k-max pooling operation

operation so that the size of input feature map of convolutional layer is the same as
the size of output feature map of the convolutional layer.
The activate function of convolutional sublayers is ReLU function:

x if x ≥ 0
φ(x) = (6.7)
0 if x < 0

Though, the performance of ReLU function is poor than sigmoid function’s, it is

simple and takes less training time to achieve convergence.
Max Pooling Sublayers Pooling sublayers are followed convolutional sublayers.
The pooling operation is also called down sampling. In pooling operation, both
width dimension and height dimension of the feature map will be reduced. However,
the depth dimension of the feature map remains unchanged. Taking 2 × 2 pooling
window for example (it can be seen in Fig. 6.3a), the width and height of feature
map will reduce half after a pooling operation. If the size of input feature map of
multi-layer network is defined, after several times pooling operation, the size of
output feature map is also definite.
The formula of convolution is as follows:

Xjl = βjl down(Xjl−1 ) + biasjl (6.8)

6.1 CNN Based Sentence Similarity Model 207

where βjl denotes the weight, down(x) denotes the down sampling, biasjl denotes
the bias.
The k-Max Pooling Sublayer The length of the sentence is variable. To support
sentences with different length, the size of input layer of the convolution neural
network must be changeable. To remain the same size of output layer of the
convolutional neural network with variable input feature map, k-max pooling is
used on the last pooling layer to replace max pooling operation. Figure 6.3b shows
k-max pooling operation (k = 3) on the input feature map. The pseudo code of
k-max pooling is as follows:

Algorithm 2 The computation process of the k-max pooling

Input: the three-dimensional feature map, tensor ∈ Rwidth×height×depth
Output: The k vectors, vectors ∈ Rk×depth ;
1: function kMax(tensor)
2: tensor = reshape(tensor, [width × height, depth])
3: //After the reshape operation, tensor is a two-dimensional matrix: tensor ∈
R(width×height)×depth
4: for i = 1 to tensor.depth do
5: tensor[:][i] = AscendingSort (tensor[:][i])
6: end for
7: for i = 1 to k do
8: vectors[i][:] = tensor[i][:]
9: end for
10: return vectors
11: end function

The k-max pooling used in our model is different from the k-max pooling used in
Nal et al.[12]. In Nal et al.[12], after k-max pooling, the relative position of data in
the original feature map is preserved. In contrast, we discard the relative position of
data because we subsequently summed the data up to calculate the sentence vector.
In Fig. 6.3b, the k-max operation on any n × n × d (n ≥ 2) input feature map
produces three d-dimensional vectors.
The Sum Function Taking the sentence “The cat sits on the mat” for example, its
subject, predicate and object convey enough meaning to imply the whole sentence.
Similarly, we hope that words or phrases that convey the meaning of sentence can be
“extracted” from the sentence by means of k-max pooling operation. The k vectors
outputted by k-max pooling is called the vector group of the sentence. There is no
fully connected layer in our network. The sentence vector is the average value of the
vector group of the sentence and is calculated by:

1
k
vS = ui (6.9)
k
i=1
208 6 Intention Based Networking Management

where vS represents the sentence vector, u1 , . . . , ui , . . . , uk denotes the vector

group of the sentence.
Algorithm 3 shows the pseudo code of the shallow convolutional neural network.

Algorithm 3 The computation process of the shallow convolutional neural network

Input: the input feature map, map0
Output: The sentence vector, vs
1: Initialize the filter: f ilter 1 , f ilter 2 , . . . , f ilter l
2: for i = 1 to l − 1 do
3: convi = ReLu(convolute(mapi−1 , f ilter i , padding = same ))
4: mapi = MaxP ooling(convi )
5: end for
6: convl = ReLu(convolute(mapl−1 , f ilter l , padding = same ))
7: vectors = kMax(convl )
8: vs = average(vectors[1], vectors[2], . . . , vectors[k])

In our application, the number of layers in our network is no more than three. The
work of fully connected layers is substituted by the next calculation unit. Simple in
structure and fast in training are the advantages of our model. The model with two
layers of network is shown in Fig. 6.4.

6.1.1.5 Similarity Computation

The similarity of sentences is calculated by the sentence vector pair [63]. There are
many methods to compute the similarity of vectors. We choose cosine distance,
Euclidean distance and Manhattan distance to evaluate the score of similarity.
However, the value range of Euclidean distance and Manhattan distance is not [0,1].
We need to make modification to them. The score calculated by Euclidean distance
and Manhattan distance is as follows:

score = f (||vs1 − vs2 ||), score ∈ [0, 1] (6.10)

where f denotes a monotone decreasing function. And the codomain of f is [0, 1].
We use three different f in the model: f1 (x) = e−x , f2 (x) = 1+e1 −x and f3 (x) =
1
1+x . Since the output of ReLu function are not negative, there is no need to concern
the negative value of cosine distance. The cosine distance of the sentences vector
pair can be directly used as the score of similarity.
6.1 CNN Based Sentence Similarity Model 209

Fig. 6.4 The sentence similarity model

6.1.2 Experimental Results and Analysis

In the subsection, we present detailed experimental setup and loss functions in our
training process, after which experimental analysis is illustrated respectively. Two
different data set SICK and MSRP are used to evaluate our model.

6.1.2.1 Experimental Setup

We did all the experiments on a personal computer with 8 gigabytes memory and
Intel i7 quad core CPU. In experiments, we use GloVe word embedding [66] (trained
on Wikipedia 2014 + Gigaword 5). And we fine-tune the word embeddings on
training. Words which not present in the GloVe are initialized randomly.
If padding is not allowed in pooling sublayer and k-max pooling sublayer, the
length of sentence must meet a lower limit. The length of the sentence satisfies:

LengthS ≥ (fw )L−1 k (6.11)

210 6 Intention Based Networking Management

where fw represents the width of pooling window (In this part, the width of pooling
window is equal to its height), k denotes the parameter of k-max, L denotes the
number of hidden layers.
In the model, the number of channels of input layer is the same as the dimension
of word embedding (commonly more than 50-dimension). Adding a new hidden
layer to network will greater increase the number of training parameters in the
model. To limit the number of parameters, the number of layers L is no more than
three.
On training task, to keep the same size of input map in the same batch, we have
to use zero padding in input layer. However, on testing task, the model supports the
different size of input map.
The model contains many hyper parameters. This is a shortcoming of the model.
There are the number of layers L, the parameter k, the width of convolutional filter,
the width of pooling window and the number of hidden feature maps to be artificially
specified. But the model with non-fully connected layer and fewer hidden layers is
less than most of deep neural networks on the total amount of parameters.

6.1.2.2 Training

The data set of our experiments are SemEval-2014 Sentences Involving Compo-
sitional Knowledge (SICK) data and Microsoft research paraphrase identification
(MSRP)[32, 35]. SemEval (Semantic Evaluation) is a computational semantic
analysis system organized by the Special Interest Group on the Lexicon of the
Association for Computational Linguistics. SICK is the data set of SemEval-2014
[36] Task 1 competition, the aim of which is evaluated compositional distributional
semantic models on full sentences through semantic relatedness and entailment.
Many participants submitted their works on the task and their results are available in
Marelli et al. [37]. SICK contains training data (4500 sentence pairs), trial data (500
sentence pairs) and test data (4927 sentence pairs). Relatedness score of sentence
pair can range from 1 (completely unrelated) to 5 (very related). MSRP contains
training data (4076 sentence pairs) and test data (1725 sentence pairs). Relatedness
score of sentence pair is one or zero.
Hyper parameters of the model were set as follows: parameter k of k-max pooling
was 3; the size of convolutional filter was 3 × 3; the single training batch size was
set to be 50. We are fine tuning the word embedding on both tasks. The loss function
of these two tasks are different.
On the SICK data set, we train model to minimize the loss function of mean
squared errors (MSE):

1 
m
loss1 = (simp − siml )2 (6.12)
m
i=1
6.1 CNN Based Sentence Similarity Model 211

where simp denotes the predicted value of similarity score, siml represents
similarity score that is calculated as the average of ten human ratings collected for
each pair, m is the scale of training data set.
Besides, to explore the influence of loss function on the result, we use the other
loss function—KL divergence loss:

1 
m
q (1 − q)
loss2 = [q · log + (1 − q) · log ] (6.13)
m p (1 − p)
i=1

where p is normalized simp and q is normalized siml . Considering denominator p

and 1 − p could be zero, Laplace Smoothing is being used on this loss function.
On the MSRP data set, we training model to minimize cross entropy between the
predicted value and the label.

6.1.2.3 Experiments Analysis

Figure 6.5 illustrates results on the SICK test set. The six results at the bottom of the
figure are the best six results released by SemEval. These results represent the high
level of traditional methods. It can be seen that the result of this model is ranked
third in above results, according to Pearson coefficient (SemEval official ranking).
Similarly, according to Spearman coefficient, the result is ranked the third. And the

Fig. 6.5 The test result on SICK data set. DT-RNN [19], SDT-RNN [19], ECNU [38], Meaning
Factory [40], UNAL-NLP [41], Illinois-LH [42], CECL [43], SemantiKLUE [45]
212 6 Intention Based Networking Management

result on MSE is ranked the fourth, very closed to the third 0.3593. There is no
distinct difference between the result of our model and the best results. Differences
between our result with the best on Pearson, spearman and MSE respectively are
0.0159, 0.0137 and 0.0382. Figure 6.5 also provides a comparison of the result
of this work with results of two different RNN models. Compared to RNN, it
is disadvantaged for the CNN network to deal with long-term and short-term
dependencies between words in the sentence. Surprisingly, the test result of the
model is slightly better than results of these two RNN models. But the results of our
model are not as good as the results of literatures (He et al. [14], He and Lin [16]).
The results of their work on Pearson coefficient are 0.8686 and 0.8784 respectively.
Despite the performance of our model is poorer than these deep neural networks, the
work of us is still valuable: Compared with traditional methods, our work achieves
better performance. And compared with deep neural networks, our model is very
practical in applications since it is simple in architecture and fast in the training.
Table 6.1 presents the methods and resources used by several models that well-
performed on the SICK data set. It can be seen that the well-performed ECNU model
uses four learning methods, WordNet and additional Corpus; the second ranked
model, The Meaning Factory, also uses three different resources including WordNet.
In contrast, our model only uses word embeddings and convolution neural networks.
Our model is much simpler in structure, but can achieve the same degree of test
performance.
Figure 6.6 demonstrates the training curve of our model on SICK set (Learning
rate is 10−3 ). From Fig. 6.6, it can be seen that the performance of these two loss
functions has no significant difference on the test results. The results show that both
curves are fast convergence.
We used different dimensional word embeddings to test our model, and recorded
their training time. We chose a single-layer network to accomplish the experiments.
Table 6.2 shows the corresponding results. All of experiments that list the training
time are done on the same computer.
As can be seen from Table 6.2, the result of a single-layer network which contains
1.62 million parameters and spent around 80 min on training can rank the third in
Fig. 6.5. After reducing the half number of parameters, the test result does not show
obvious difference with the same epoches. The Pearson correlation coefficient of the
result reaches 0.793, the fourth level in Fig. 6.5, and are very closed to that of the
third. Meanwhile, the consuming time can be reduced to a half: the training process
can be accomplished in 37 min.
Though the model of the third line in the table (100-dimensional word vector)
has not bright performance under ten epoches, if the epoches are 40, the metrics of
its Pearson correlation coefficient was 0.7868 around, its mean squared error was
approximately 0.3878, while the parameters of the whole model are only 3/8 of the
second case in the table.
Figure 6.7 shows that the result of our model which is pretrained on SICK
corpus outperform others on MSRP test set. Meanwhile, the result of our model
without pretraining does not perform well. Thus, pretraining is helpful to improve
the performance of our model. We assume that our model may perform better after
Table 6.1 The methods and resources that models applied [37]
Learning methods External resources
SVM and kernel K-nearest Classifier Random Convolutional Paraphrases Other STS MSR-video Word
Model methods neighbours combination forest neural network Other WordNet DB corpora ImageFlicker description embeddings
6.1 CNN Based Sentence Similarity Model

ECNU Y Y Y Y Y Y
The Meaning Y Y Y Y
Factory
UNAL-NLP Y Y Y Y
Illinois-LH Y Y Y Y
This work Y Y
Y represents the model used this method/resource
213
214 6 Intention Based Networking Management

Fig. 6.6 The relatedness curve of mean square errors, Pearson coefficient and epoches on test data
set using KL divergence loss function and MSE loss function

pretraining on a large amounts of appropriate training data, which we leave to our

future work. (In Fig. 6.7, the work of Hu et al.[6] is based on CNN, others’ work in
Fig. 6.7 are based on traditional methods. The results of Baseline and Rus et al.[46]
is reported by Hu et al.[6], and the result of Rus et al.[46] is the best one in the
report.)
6.2 A Feature Selection Method Based Text Classification System 215

Table 6.2 The results on the params r ρ MSE T

different settings of the model
1,620,002 (300) 0.8069 0.7433 0.3897 4731
720,002 (200) 0.7930 0.7308 0.4269 2197
270,002 (100) 0.7679 0.7052 0.4419 819
90,002 (50) 0.7623 0.6999 0.5076 495
The “params” represents the parameters of the whole
model to be trained, the content in the parentheses
denotes the dimension of word vector. r denotes Pearson
correlation coefficient, ρ denotes spearman correlation
coefficient, MSE denotes mean square error, and T
denotes the average time spent for the training (Unit:
seconds). The epoches of the results are ten

Fig. 6.7 The test result on MSRP data set Baseline[6], Hu et al.[6], Rus et al.[46], Blacoe et
al.[47], Fernando and Stevenson [49]

6.2 A Feature Selection Method Based Text Classification

System

With the development of the Internet, the amount of Chinese text information shows
an exponential growth trend [50]. How to effectively manage the massive Chinese
documents and mine the information contained in the documents has become a
critical research problem. Automatic Text Classification can complete the work
of text processing effectively. It also plays an important role in natural language
processing (NLP) and data mining.
216 6 Intention Based Networking Management

Train Train Test

Data

Test Feature
Proprocess
Data Selecon

Weight Synonym
Calculaon Merging

Train Text Text

Evaluaon
Classificaon Classifier

Fig. 6.8 Text classification model framework

The most common method used in text classification is the vector space model
(VSM). It represents text as a feature vector. The specific process is shown in
Fig. 6.8.
From the above figure, we know that the first step in Chinese text classification is
to preprocess the text, including word segmentation, part of speech tagging, removal
of stop words, etc. The purpose is to remove the useless words, and only leave the
nouns, adjectives and verbs that contain category information. After this, the text can
be represented as a vector to form VSM. Then we use the feature selection method
to select the feature words that can symbolize the text categories, and merge the
synonym to reduce dimensions. Next, TF-IDF [9, 18] method is used to calculate
the weight of each feature of each text to transform the text into a feature vector.
Last but not least, by using the Bayesian classifier to train the sample data, we can
get the final text classifier.
Feature selection is the most important step because the selected feature words
directly affect the accuracy of the classifier. In VSM, the best feature selection
method is χ 2 statistics (CHI) [21, 52, 53]. But the defect is high-dimensional feature
vectors selected by CHI may cause dimension disaster. The writer consider to merge
the synonyms among the feature words selected by CHI so that the dimension of
feature space can be reduced. Then in next step an improved TF-IDF method is
used to calculate the feature weights for each word to generate the feature vector of
each text. This part mainly studies the influence of feature selection and synonym
merging on the accuracy of classification in automatic text classification. Synonym
merging can reduce the dimension of the feature space and improve the classification
6.2 A Feature Selection Method Based Text Classification System 217

performance. The study of feature selection algorithm and synonym merging has a
strong practical significance. The main contributions of this part are as follows:
1. We presented a new feature selection algorithm named SM-CHI based on an
improved CHI [22, 54] formula and synonym merging to achieve efficient feature
selection and dimension reduction;
2. We found that the original CHI formula multiplied by a log term has the best
classification performance comparing with the original CHI and two improved
CHI algorithms [28, 54, 55];
3. The choice of thresholds α (0  α  1) is critical. Only the most similar feature
words will be merged when α is close to 1, so we use grid search method to find
the optimal α. The result show that the classification accuracy is highest when α
is equal to 0.8;
4. We proposed three improved weight calculation methods based on TF-IDF. The
experimental results show that using the maximum value of the synonym group
as the feature weight is the best way.

6.2.1 Text Classification Model Based on Semantic Similarity

In this subsection, we mainly introduce the text classification model based on

semantic similarity. Wherein, Sect. 6.2.1.1 describes the feature selection method
based on χ 2 statistic. Section 6.2.1.2 introduces the method of synonym merging;
Sect. 6.2.1.3 presents the traditional weight calculation method, TF-IDF.

6.2.1.1 Improved Feature Selection Algorithm

In text classification, a feature word and its category tend to obey the CHI formula.
Higher CHI value implies that a feature word has stronger ability to identify a
category. The CHI value of word is calculated as follows [53]:

N ∗ (AD − BC)2
χ 2 (t, c) = (6.14)
(A + C)(A + B)(B + D)(C + D)

where N is the size of the training set, A is the number of documents that belong to
class c and contain the word t; B is the number of documents that do not belong to
class c but contain the word t; C is the number of documents that belong to the class
c but do not contain the word t; D is the number of documents that do not belong to
class c and don’t contain the word t.
Although the CHI formula has a relatively good performance in text classifica-
tion, it also has some shortcomings [24, 59]. First of all, high-frequency words that
appear in all categories have higher CHI values, but they do not make much sense
for class distinctions. Secondly, the CHI formula only considers the appearance of a
218 6 Intention Based Networking Management

word but not the frequency of the word in a document. Therefore, CHI formula also
has “low frequency words flawed”. For example, assuming word t1 appears in 99
documents, each appears ten times; word t2 appears in 100 documents, each appears
one times; obviously t2 has a higher CHI value, but in fact t1 is more representative
for this category. There are many studies amended for its defects. The work in [54]
proposed the multiplication by a log entry based on the original CHI to reduce the
CHI value of high-frequency words. The formula is as follows:

N
chi_imp_1 = log( ) ∗ χ 2 (t, c) (6.15)
A+B

where A + B represents the number of documents that contain word t, and N

represents the total number of documents. In this case, the CHI value of the high-
frequency words that appear in all categories are close to zero so that they won’t be
selected as a feature word.
In addition, the work in [55] has made a corresponding improvement to the word
frequency, which is multiplied by term β(t, c) on the basis of the original CHI
formula. The calculation formula is as follows:

chi_imp_2 = β(t, c)χ 2 (t, c) (6.16)

where β(t, c) is calculated as follows:

tf (t, c)
β(t, c) = m (6.17)
i=1 f (t, ci )

In the formula, m is the total number of categories, tf (t, c) is the frequency of

the word t in the category c.

6.2.1.2 Synonym Merging Algorithm Based on “Tong YiCi Cilin”

Some of the feature words selected by CHI formula may be the same or have
similar meaning. They have the same effect on class distinctions. If the synonym
are merged, not only the classification accuracy will be improved, but also the
dimension of the feature space can be reduced so that the efficiency of the algorithm
can be improved. For example, “GanMao”, “ZhaoLiang”, “ShangFeng” are the
synonym of “Cold” in Chinese. If the “Health Care” category articles contain
these words respectively, then the feature words for the text classification contain
too much redundant information. We use the method of synonym merging to deal
with it. In this part, we use the “Tong YiCi Cilin” provided by Harbin Institute of
Technology as the method of word similarity calculation [60]. Its structure has five
layers. You can easily calculate the similarity between the two terms. The structure
of “Tong YiCi Cilin” is shown in Fig. 6.9.
6.2 A Feature Selection Method Based Text Classification System 219

First layer

Second layer

Third layer

Forth layer

Fifth layer

Fig. 6.9 Hierarchy structure of “Tong YiCi Cilin”

The concrete similarity calculation method is introduced in detail in [60]. When

the similarity of two words is greater than threshold α, they will be regarded as a
pair of synonym to merge. The optimal value of α will be discussed later in the
experiment. In addition, all merged synonyms are stored in a list. Nested lists are
used to store feature words so that all synonym information remains in the feature
vector. To calculate the feature vector of each document, we propose three improved
methods which will be discussed in Sect. 6.2.1.3.

6.2.1.3 Weight Calculation Method

Traditional TF-IDF weight calculation formula [9] is as follows:

weightt,d = tft,d + idft (6.18)

This formula represents the weight calculation method for word t in document
d. Here, tft,d denotes the frequency of occurrence of word t in d, and idft denotes
the anti-document frequency of t, which is used to quantify the distribution of t in
the training set. If n is used to denote the number of documents which contain t in
the training set, the calculation formula of idft is as follows:

N
idft = log( ) (6.19)
n
As we mentioned above, the TF-IDF method is used to calculate the weight of
each feature word in each text.
220 6 Intention Based Networking Management

6.2.2 Algorithm Description

6.2.2.1 Feature Selection Method Based on the Synonym Merging

In this subsection, we introduce the feature selection algorithm (SM-CHI). This

method firstly selects candidate feature words based on an improved CHI formula,
and then merge synonyms to re-select those feature words that can represent the
categories better and reduce dimension. The method is represented as the following
formula:

SM − CH I = LF (t) ∗ CH I (t) ∗ SM(t) (6.20)

where LF (t) denotes whether the word t exists in the word bag or not, and is
mainly decided according to the part of speech and stopping words. If the word
t is a stopping word and in the part of speech that does not belong to verb, noun,
adjective, LF (t) = 0, otherwise LF (t) = 1. CH I (t) represents the CHI value
of the word t, and is calculated by Eq. (6.15). SM(t) indicates whether the word t
contains synonym. If yes, it needs to merge all of its synonyms.
Firstly, all the texts in the training set are preprocessed, including Chinese word
segmentation, part-of-speech tagging, and discarding stop words. The remaining
words constitute the word bag of the training set. Secondly, we calculate the CHI
value of each word. Choose the first 200 words from each category to form candidate
sets of feature words. Note that the characteristic words selected for each category
may be duplicated. The candidate set is stored using the HashSet (a data structure)
and the de-emphasis is performed. After obtaining the candidate set, the similarity
between each word is calculated according to “Tong YiCi Cilin” and threshold is set
to α. The synonym merging is performed only when the word similarity is greater
than α. We will experimentally determine the optimal value of hyper-parameter α.
The pseudocode of SM-CHI is shown in Algorithm 4.

6.2.2.2 Improved Method for Calculating Eigenvalue Weight

In the scene of SM-CHI feature selection method presented in this part, the
traditional TF-IDF formula has some drawbacks. For the features after synonym
merging, the original weight calculation formula will cause “unfairness”. Because
the merged feature words are stored in the nested list, so which word among them
will be regarded as the feature is a question. For this problem, we present the
following three solutions:
• Sum the weights of all items up in the feature list of each dimension as the weight
of the list;
• Take the largest weight among the synonym as the weight value of the feature;
• Multiply the first item by 1.1 for times of the number of items in the feature list.
6.2 A Feature Selection Method Based Text Classification System 221

Algorithm 4 SM-CHI feature selection algorithm

1: Input:a training set D
2: Output:a feature space F
3: (Initialization) A,B,T F I DF can all be a null dict
4: for each file in D:
5: word_list=file.process()
6: for word in word_list:
7: A[file.class][word] += 1
8: T F I DF [file.class][file.num][word] +=1
9: end for
10: Calculate B in the CHI formula from A
11: for cla in
12: for word in cla:
13: Calculate the CHI value according to formula (6.15)
14: Selects the first 200 words as the feature
15: end for
16: Combine features of the 9 categories to obtain word_features
17: for word1,word2 in word_features:
18: sim=calcWordsSimilarity(word1,word2)
19: if sim > α:
20: Merge word1 and word2
21: end for

6.2.3 Experiments and Results

In this subsection, the following three groups of experiments are carried out to test
the three innovation points of SM-CHI with control variable method. In Experiment
I, we test the three feature selection algorithms without using synonym merges. In
Experiment II, we use the first improved CHI method to select features and use grid
search method to find the optimal threshold α. On the basis of Experiment I and II,
we designed Experiment III to find the best weight update method.

6.2.3.1 Performance Evaluation and Data Set

The standard precision rate P , recall rate R and F 1 score are used to measure the
classification performance. For the i_th category, the formula is as follows [62]:

TP TP 2 ∗ Pi ∗ Ri
Pi = , Ri = , F 1i = (6.21)
T P + FN T P + FP Pi + Ri

where T P is the number of documents correctly classified as class i, F P is the

number of documents classified as class i but not actually i, and F N is the number
of documents that is not classified as class i but is actually class i.
This article will use the whole network news data set provided by Sogou Lab
to test our experiments. The corpus includes nine kinds of news types, such as
Automobile, Finance, IT, etc. Each category contains thousands of documents. In
222 6 Intention Based Networking Management

this experiment, each category takes 400 documents, of which 280 are training set
and 120 are test sets. Therefore, the training set contains 2520 documents, and the
test set includes a total of 1080 documents.
The preprocessing module uses a third-party library for python, named jieba, to
complete the work of word segmentation, part of speech tagging and discarding
of stop words. In addition, we use Naive Bayesian classifier provided in Python’s
NLTK library as the classifier.

6.2.3.2 Experiments and Results

In this section, the following three groups of experiments are carried out to test the
three innovation points of SM-CHI with control variable method. In Experiment I,
we test the three feature selection algorithms without using synonym merges. In
Experiment II, we use the first improved CHI method to select features and use grid
search method to find the optimal threshold α. On the basis of Experiments I and II,
we designed Experiment III to find the best weight update method.

Experiment I

In order to test the effect of three kinds of feature selection methods described in
Sect. 6.2.1, we conducted the following experiments. But we didn’t use synonym
merging here. The results of experiment are shown in Fig. 6.10.
From the results, we can see that the two improved CHI formulas have a great
effect on enhancing the value of F1 score of each category as compared to the
original CHI formula, which means that the improved CHI formulas can select more
representative words. They both make some improvement based on the original
CHI. In addition, when the two improved CHI formulas are compared, the first
improved method has a slight advantage, showing a better discrimination effect
in the preceding categories. The result also shows that the log term successfully
suppresses the CHI values of the high frequency words appearing in all classes,
which achieves relatively good results. Therefore, we will use the first improved
CHI formula as our base feature selection method in the fellow experiment.

Experiment II

In order to select the appropriate threshold α for synonymy merging, we designed

the following experiment. The first improved CHI formula was used for feature
selection, and the range of α is set to [0.5, 0.6, 0.7, 0.75, 0.8, 0.85, 0.9,1.0]. A
total of nine experiments are conducted, including a comparative experiment that
didn’t use synonym merging. The experimental results are shown as Table 6.3 and
Fig. 6.12.
6.2 A Feature Selection Method Based Text Classification System 223

Fig. 6.10 Classification results comparing three kinds CHI formula

Table 6.3 F1 score of each category when α takes different values

Car Finance IT Health Sports Tourism Edu Culture Mil
α = 0.50 0.729 0.701 0.579 0.833 0.769 0.805 0.704 0.693 0.900
α = 0.60 0.747 0.787 0.709 0.872 0.815 0.810 0.712 0.713 0.895
α = 0.70 0.734 0.814 0.803 0.874 0.919 0.817 0.694 0.764 0.896
α = 0.75 0.757 0.813 0.825 0.861 0.946 0.808 0.726 0.779 0.900
α = 0.80 0.763 0.821 0.828 0.931 0.949 0.827 0.726 0.819 0.903
α = 0.85 0.763 0.812 0.818 0.931 0.937 0.821 0.721 0.804 0.889
α = 0.90 0.782 0.783 0.808 0.943 0.945 0.781 0.720 0.802 0.913
α = 1.0 0.782 0.781 0.788 0.932 0.936 0.789 0.714 0.794 0.897
None 0.782 0.776 0.783 0.935 0.936 0.778 0.711 0.788 0.901
Bold values indicate the maximum value in each category

From the results in Figs. 6.11 and 6.12, we can draw the conclusion that the
classification accuracy is the highest when α = 0.8 and worst when α = 0.5.
The use of synonym merging improved classification accuracy by approximately 3
percentage points compared to use CHI only. By specific analysis of each category,
we found that when we use synonym merging only the first category has a low
accuracy compared to no synonym merging. The reason is that the eigenvectors
after synonym merging have reduced the text discrimination degree of the “car”
category.
224 6 Intention Based Networking Management

Fig. 6.11 F1 score of each category when α takes different values

Fig. 6.12 Total accuracy when α takes different values

Experiment III

Based on the previous two experiments, we designed the following experiment to

select the optimal weight update method introduced in Sect. 3.3. We designed three
experiments, all with the first improved CHI formula and a threshold of 0.8. The
experimental results are shown in Fig. 6.13:
6.3 Sematic and Statistical Information Based Text Classification 225

Fig. 6.13 Three kinds of special effects classification weight value updating method of compari-
son

According to Fig. 6.13, it can be seen that the classification accuracy of method
1 is the lowest. The reason is that this method adds the weights of all the synonymy
words as the weight of the feature, but a word and its synonym words appear in more
than one category. This simple superposition will make the feature differentiate
the category worse. In contrast, methods 2 and 3 use the combined synonym as
a one-dimensional feature and achieve better classification results and F1 scores. In
contrast, method 2 is more effective which shows that the maximum value of the
synonym is a better method because it can represent the maximum abilities of all
synonyms to differentiate the text categories. By multiplying the power of 1.1 by
the n, method 3 incorrectly increases the ability of the feature to distinguish text
categories.

6.3 Sematic and Statistical Information Based Text

Classification

In recent years, with an increasing volume of text information on the Internet and
social media, text categorization has become a key technique to process these textual
data. In text categorization, a Bag of Words (BOW) is usually used to represent a
document. The weight in BOW is usually obtained by computing word frequency or
the widely accepted TF-IDF formula. However, the BOW representation has several
226 6 Intention Based Networking Management

limitations: (1) The TF-IDF formula does not consider the calculation of class-based
weights. As a result, the same word has the same weight in all categories. (2) It
cannot deal with synonyms and polysemy.
In the absence of knowledge-based word similarity and statistic-based word sim-
ilarity, automatic text categorization using BOW only as a document representation
model[64] has not yet achieved the best performance and cannot meet the needs
of all scenes in the real life. There are two ways to address this problem. First,
we could use language model based on deep learning models such as word2vec
[65] and Glove [29, 66] to learn the vector representations of words. However,
such new approaches do not have to be necessarily better when the corpora is
not particularly large. And it takes considerable time and effort to train word
vectors. The second method is to collect semantic and syntactic informations as
much as possible. We mainly adopt the second method and develop a new semantic
smoothing kernel function based on knowledge-based word similarity and statistic-
based word similarity to increase the capability of feature vectors to represent a
document [33].
This chapter presents a novel approach for text classification. In this approach,
word similarity based on HowNet is embedded into semantic information. This
method promisingly improves the accuracy of text classification via using ontology
knowledges. Moreover, the proposed approach takes advantage of the class-based
term weighting by giving more weights on core words in each class during the
transformation phase of SVM from the input space to the feature space. A term
has a more discriminative power on a class if it has a higher weight for that class.
The heuristic idea combining sematic and statistical information finally improves
the classification accuracy.

6.3.1 Word Similarity

6.3.1.1 Support Vector Machines for Classification

Support vector machine (SVM) is a very effective machine learning algorithm

developed from statistical learning theory. This algorithm was proposed by Vapnik,
Guyon and Boser [67] and further analyzed in [68]. The core goal of SVM is to find
the optimal segmentation hyperplane by the maximum spacing between classes. The
author of [69] proposed that SVM has many advantages, such as finding the global
optimal solution and having a good robustness.
SVM kernel function can be regarded as similarity function, because it calculates
the similarity values of data sets. It is proposed to define a suitable kernel function
[51, 70], which has a direct influence on finding the optimal hyperplane. The
commonly used kernel functions for the document vectors are given below:

LinearKernel : k(dp , dq ) = dp dq (6.22)

6.3 Sematic and Statistical Information Based Text Classification 227

P olynomialKernel : k(dp , dq ) = (dp dq )b , b = 1, 2, . . . (6.23)

RBF Kernel : (dp , dq ) = exp(γ ||dp − dq ||)2 (6.24)

In current works, the authors of [71] proposed to develop a kernel function

based on the similarity of the knowledge system, and used the Omiotis library
function to measure the similarity of English words and improve the accuracy of
the classification. The authors of [72] proposed to develop a kernel function based
on the weights of the class which improved the accuracy of the classifier. Based on
these research efforts, this part optimizes the similarity of Chinese text words and
combines the statistical methods with the knowledge-based methods to construct a
kernel functions to improve the accuracy of text classification.

6.3.1.2 Knowledge-Based Word Similarity

Knowledge-based systems use ontology or thesaurus to capture the concepts in

the documents and incorporate the domain knowledge into the words for the
representation of textual data. These systems enhance the representation of terms
by taking advantages of semantic relatedness among terms.
The similarity calculation of words is added to the text classification [34, 73],
which is used to modify the weights of the text feature vectors. Mavroeidis et al. [74]
proposed a semantic kernel function based on WordNet [75] to improve the accuracy
of English text classification. Based on the Chinese semantic knowledge system
of HowNet [76], Zhang embedded the semantic similarity into the kernel function
of Chinese text classification [77], and improved the performance of Chinese text
classification.
In this part, we use the Chinese dictionary HowNet to calculate the semantic
similarity of words. HowNet is a very detailed dictionary of semantic knowledge.
Unlike CiLin [78] and WordNet, every word in HowNet has multidimensional
knowledge representations. The structure of HowNet is described in detail below.
HowNet mainly includes concepts and primitives. Each term is described by a
number of concepts, each of which is described by a sequence of primitives, so
primitive is the smallest expression unit in HowNet. HowNet contains 1500 prim-
itives, which can be divided into three categories: basic semantics (describing the
semantic features of concepts), grammatical semantics (describing the grammatical
features of words) and relational semantics (describing the relationship between
concepts). When we calculate the word similarity, we can define it in the following
way. Word similarity calculation consists of four parts in Eq. (6.25).


4 
i
Sim(S1 , S2 ) = βi Simj (S1 , S2 ) (6.25)
i=1 j =1
228 6 Intention Based Networking Management

Sim1 (S1 , S2 ) is the similarity of first basic primitives of these two words. The
similarity Sim1 (S1 , S2 ) between S1 and S2 can be calculated using Eq. (6.26).
Sim2 (S1 , S2 ) is the similarity of the rest basic primitives, that is the arithmetic
mean of the similarity of all pairs of elements. Sim3 (S1 , S2 ) is the similarity of two
grammatical semantics, which can be transformed into the basic semantic meaning
in the grammatical semantics. Sim4 (S1 , S2 ) is the similarity of two relational
semantics, but the elements in the relational semantics are sets, which are basic
primitives or concrete words.
There is a close relationship between word similarity and word distance. In fact,
word similarity and word distance are different forms of the same feature of a pair
of words. Word similarity is defined as a real number between 0 and 1.
α
Sim1 (S1 , S2 ) = (6.26)
d +α

where S1 and S2 represent two of the words respectively, d is the distance between S1
and S2 in the original path hierarchy in HowNet; α is an adjustable parameter. When
the distance in HowNet between words is particularly large, Sim(S1 , S2 ) approaches
0; when the distance in HowNet between words is particularly small, Sim(S1 , S2 )
approaches 1.
βi in Eq. (6.25) is an adjustable parameters and satisfies Eq. (6.27). The lat-
ter part of Eq. (6.27) represents the descending importance of Sim1 (S1 , S2 ) to
Sim4 (S1 , S2 ).

β1 + β2 + β3 + β4 = 1, β1 ≥ β2 ≥ β3 ≥ β4 (6.27)

6.3.1.3 Statistic-Based Word Similarity

In the absence of semantic knowledge, a statistical-based approach such as the

one presented in [39, 79] can be applied to text categorization to solve synonymic
problems. Statistical similarity calculation is based on the correlation of training
words, so the statistical similarity calculation method is very sensitive to the training
data sets.
The similarity calculation using statistics in text categorization contains calcu-
lation based on classes [72], high-order paths [80–82] and mean value calculations
[83]. In this part, we use the method proposed in [72] which considers the weight of
a certain word in a training set depending on the relevance of terms and categories.
After the text feature vector is smoothed through the semantic kernel, it can
improve the weights of important words in the category and reduce the weight of
common words in the category. By modifying weights of the text feature vectors,
the representation capability of the feature vectors is increased.
6.3 Sematic and Statistical Information Based Text Classification 229

6.3.1.4 Weight of Feature Words

In the classification system, the most commonly used method of calculating word
weights is the TF-IDF formula mentioned in [44, 58, 85, 86] where TF denotes
the term frequency and IDF denotes inverse document frequency. TF-IDF formula
was first used in the field of information retrieval, because its calculation method is
simple and practical. It is also widely used in the text automatic classification.
TF-IDF is a statistical method to evaluate the importance of a document in a
corpus. In general, the importance of a word increases in proportion to its number
of occurrences in the document and decreases inversely with its higher frequency of
occurrences in the corpus.
IDF formula is given in Eq. (6.28)

|D|
I DF (w) = (6.28)
dfw

where |D| denotes the total number of documents in the corpus and dfw
represents the number of documents which contains term w.
TF-IDF formula is given in Eq. (6.29)

T F I DF (w, di ) = tfw ∗ log(I DF (w)) (6.29)

where tfw represents the term frequency which is the number of word w in
document D.
TF-ICF is proposed as another method of calculating word weight in [87, 88],
which is similar to TF-IDF. ICF denotes inverse class frequency. TF-ICF calculates
the word weight in category level rather than in document level.
Equation (6.30) shows the ICF formula:

|C|
I CF (w) = (6.30)
cfw

where |C| denotes the total number of classes in the corpus and cfw represents the
number of classes which contains term w.
TF-ICF formula is shown in Eq. (6.31),

T F I CF (w, ci ) = tfw ∗ log(I CF (w)) (6.31)
d∈cj

Inspired by the IDF and ICF formulas, [48, 89, 90] proposes a new method for
calculating weights:

|D|
Ww,c = log(tf cw,c + 1) ∗ (6.32)
dfw
230 6 Intention Based Networking Management

where tf cw,c represents the total number of feature term w of class c. |D| denotes
the total number of documents and dfw represents the number of documents which
contain term w.
From analysis above, we can see that W is a matrix which is determined by
categories and feature terms. In fact, terms that are similar to the topic in the
category are given a larger weight because of the W matrix. The authors of [89, 90]
compare the weighting algorithm based on the category with other commonly used
feature selection algorithms, and conclude that the former one can improve the
classification performance significantly.

6.3.2 Merged Kernel Function

Both knowledge-based word similarity computation and statistical-based word

similarity computation can improve the performance of text classifiers [84]. A
heuristic idea is to apply two computation methods to a kernel function which
we call merged kernel function in this part. The pseudocode for the merged kernel
function is shown in Algorithm 5. This section will explain in detail how to combine
the two kernel functions to improve the accuracy of classification.

6.3.2.1 Vector Space Model

Document representation is a basic problem in natural language processing. Com-

puter cannot directly deal with document which is mainly consisted of unstructured
data. The key challenge is how to map a document into a vector space model (VSM).
First, a document di is represented as an n-dimensional vector composed of feature
words as shown in Eq. (6.33).

dj = (w1j , w2j , . . . , wnj ) (6.33)

Then, the weighting formula maps the document vector di to the word weight
vector φ(dj ):

φ(dj ) = [tf idf (t1 , dj ), tf idf (t2 , dj ), . . . , tf idf (tn , dj )] (6.34)

where tf idf (ti , dj ) denotes the TF-IDF value of the feature word ti in the document
dj .

6.3.2.2 Statistic-Based Similarity Matrix

The training phase for statistic-based similarity matrix is shown in Algorithm 5

[need to rewrite, just a figure] from the first line to the fourteenth line. In order
6.3 Sematic and Statistical Information Based Text Classification 231

Algorithm 5 Calculation of the combined semantic smoothing matrix C

1: Require: Training set D
2: Ensure: Sematic Smoothing Matrix C
3: Local variables:
4: tf cω,k : total term frequency of word ω in the documents of class k
5: tfω,d : total term frequency of word ω in the document d
6: N : total number of documents in the training set
7: Nω : shows total number of documents in the training set those contain word w
8: Z: matrix is constructed according to the list of feature word by the introduction of
HowNet
9: for each word w do
10: for each document d contains word w in the train set do
11: Nω = Nω + 1
12: end for
13: end for
14:
15: for each word w do
16: for each document d contains word w in class k do
17: tf cω,k = tf cω,k + tfω,d
18: end for
19: Wω,k = (log(tf cω,k ) + 1) ∗ log(N/Nω )
20: end for
21:
22: S = W ∗ W T
23:
24: Z 2 = Z ∗ Z T
25:
26: for i in columns of S do
27: for j in rows of S do
28: Ci,j = λ1 ∗ Sij + λ2 ∗ Zij2
29: end for
30: end for

to embed the statistical information into the space vector model and increase the
capability of representing a document of feature vectors, we construct a matrix S
based on the class-based weight, which is called the statistical similarity matrix. The
formula has been described in detail in Sect. 6.3.1.4. To make use of this formula,
we define the statistical similarity matrix S as:

S = WWT (6.35)

where W is the class-based weighting formula mentioned previously. The statistical

similarity matrix S is a symmetric matrix, and the element Sij in matrix S is the
class-based statistical similarity of the feature words wi and wj .
The S matrix represents the statistical similarity of words. For example, the words
“patient” and “sufferer” have similar meaning. When the weight of the “patient” is
higher and the weight of the “sufferer” is lower in vector ϕ(dj ), the weight of the
word sufferer will be increased after multiplication with the S matrix.
232 6 Intention Based Networking Management

6.3.2.3 Combined Semantic Smoothing Matrices

The knowledge-based similarity matrix Z can be constructed according to the

list of feature words by the introduction of HowNet. Zij denotes the knowledge-
based similarity between the feature words wi and wj . The knowledge-based word
similarity can be embedded into the space vector model by matrix Z. In order to
ensure the validity of final merged kernel, a second-order rule similarity matrix is
used here, that is Z 2 = ZZ T .
After computing the statistic-based similarity matrix S and the second-order
knowledge-based similarity matrix Z 2 , the degree of weight modification of the
two matrices to the space vector model cannot be determined, which should be
determined according to the data set. In this part, the matrix S and Z 2 are combined
to generate a new semantic smoothing matrix C:

Cij = λ1 ∗ Sij + λ2 ∗ Zij2 (6.36)

where Sij and Zij2 are described in Sects. 6.3.2.2 and 6.3.2.3, λ1 and λ2 adjust the
normalization parameters of the weights in S and Z 2 . Parameters λ1 and λ2 satisfy
λ1 + λ2 = 1. We can adjust λ1 and λ2 to determine how matrices S and Z 2 affect the
classification performance of the classifier.

6.3.2.4 Semantic Smoothing Kernel Function

By defining the mapping architecture matrix C, we can map a document vector to a

new feature space vector by using Eq. (6.37).

φ(dj ) = φ(dj )C (6.37)

The mapped vectors can be directly used in many classification methods. If the
high-dimensional sparse matrix appears in the text classification, there will be a
high-dimensional disaster in computation. Defining a kernel function can reduce
the influence of the high-dimensional sparse matrix. The inner product between
documents p and q in the feature space is computed by the kernel function using
Eq. (6.38).

KCK (dp , dq ) =< φ(dp ), φ(dq ) >= φ(dp )CC T φ(dq )T (6.38)

where KCK (dp , dq ) denotes the similarity of document dp and document dq . φ(dp )
and φ(dq ) are the new feature space vectors of document dp and document dq after
transformed by the semantic smoothing matrix proposed in Eq. (6.37). The kernel
function information is stored in the matrix G:

Gp,q = KCK (dp , dq ) (6.39)

6.3 Sematic and Statistical Information Based Text Classification 233

Then we prove the validity of the semantic smooth kernel proposed in this
section. According to mercer theorem [91], any semi-definite function can be used
as a kernel function. The semantic smoothing matrix C proposed in this part is
composed of the statistical similarity matrix S and the second-order knowledge-
based similarity matrix Z 2 , so the matrix C is also a symmetric matrix. The matrix
S and the matrix Z 2 are the product of a matrix and its transpositions, so the
matrix S and Z 2 are both semi-definite matrix, which is proved in [92]. In linear
algebra, the sum of two positive semi-definite matrices is also a semi-definite matrix.
Therefore, the matrix C is a semi-definite matrix, satisfying the conditions required
by Mercer’s theorem. The kernel function can be constructed by the semantic
smoothing matrix C.

6.3.3 Experiments and Results

6.3.3.1 Corpora

This section selects the corpus provided by Sogou Company.1 and Fudan Uni-
versity.2 The Sogou corpus consists of SogouCA and SogouCS news corpora
containing various categories of 2,909,551 news articles, of which about 2,644,110
articles contain both a title and relevant content. We manually categorize articles
by using the channel in URL, then we get a large Chinese corpus with the article
contents and categorizes. However, there are some categorizes that contain few
articles. So five categories with largest number—“sports”, “finance”, “entertain-
ment”, “automobile” and “technology” are finally selected for our text classification
experiments. The details of the Sogou corpus are presented in Table 6.4. During the
training process, many parameters in the model are involved. In order to determine
the optimal values of parameters in this model, we use the validation set. We
partition training set, validation set and test set in 8:1:1 proportions in Sogou corpora
after we shuffled the corpora. So the corpora is randomly divided into training set,
validation set and test set. These parameters are described in detail in Sect. 6.3.3.4.

Table 6.4 Sogou news corpora

Category Total Train Validation Test
Sports 645,931 80,000 10,000 10,000
Finance 315,551 80,000 10,000 10,000
Entertainment 160,409 80,000 10,000 10,000
Automobile 167,647 80,000 10,000 10,000
Technology 188,111 80,000 10,000 10,000

1 http://www.sogou.com/labs/resource/list_news.php.
2 http://www.nlpir.org/download/tc-corpus-answer.rar.
234 6 Intention Based Networking Management

Table 6.5 Fudan University Category Total Train Validation Test

corpora
Economy 1589 700 100 100
Sports 1188 700 100 100
Environment 1022 700 100 100
Politics 1013 700 100 100
Agriculture 992 700 100 100

To validate the combine kernel’s effect on a small corpora, we also use the
corpus provided by Fudan University. The corpora contains 9804 articles that have
been already divided into 20 categories. We choose five categories—“economy”,
“sports”, “environment”, “politics” and “agriculture”. The details of the Fudan
training set are presented in Table 6.5. We partition training set, validation set and
test set in 7:1:1 proportions in Fudan corpus after we shuffled the corpora.

6.3.3.2 Word Segmentation and Stop Words

The current English word segmentation tool has been well developed, while the
Chinese word segmentation technology is still evolving. For python language,
NLTK [93] nltk.tokenize module can be used for word segmentation in English,
and jieba tool can be used for word segmentation in Chinese.
In order to save storage and improve the efficiency of classification, the classifica-
tion system will ignore certain words after word classification, which are called stop
words.3 There are two kinds of stop words: the first one can be found everywhere
in all kinds of documents with which the classification system cannot guarantee the
true classification result. The second kind of stop words includes the modal particle,
adverb, preposition, conjunction and so on.

6.3.3.3 Feature Word Extraction

In the problem of text categorization, a certain feature word and its class obey the
CHI square distribution. The larger the CHI value is, the more the CHI value can be
used to identify the category. The CHI formula is given:

N(AD − BC)2
χ 2 (t, c) = (6.40)
(A + C)(A + B)(B + D)(C + D)

where N is the number of texts in the training set, A is the number of documents
belonging to class c and contains the word w; B is the number of documents that do
not belong to class c but contains word w; C is the number of documents belonging

3 https://github.com/Irvinglove/Chinese_stop_words/blob/master/stopwords.txt.
6.3 Sematic and Statistical Information Based Text Classification 235

to c Class, but does not contain the word w; D is the number of documents that do
not belong to c class and do not contain the word w.

6.3.3.4 Experiment Settings

The classifier uses the SVM function provided in the machine learning library
sklearn [94] in python environment. We change the kernel function by using the
interface it provides. We observe how the statistical similarity matrix S and the
second-order knowledge-based similarity matrix Z 2 affect the performance of the
classifier when the ratio of training set and parameter λ are different.
In the experiment, we set parameter values based on the experience gained
from the validation set. First, we describe several parameters mentioned above. We
compute word similarity using α 1.6, β1 0.5, β2 0.2, β3 0.17 and β4 0.13. Second,
the length of VSM used for representing Sogou corpus is 10,000, and Fudan corpus
is 1000. Sogou corpus is large, so the feature vectors need to be longer. Finally, we
set some parameters of the model. Penalty parameter C of the error term is 1.0 and
there is no hard limit on iterations within solver, so that the SVM algorithm will
stop training before over-fitting.
A number of parameters have been used to assess the performance of classifica-
tion model output, such as accuracy [95] and f-measure (F1) [96]. To demonstrate
that the combined kernel does improve the accuracy and F1 value of text classifica-
tion, we use other machine learning methods for comparison, including KNN, Naive
Bayes, and SVM with linear kernel and RBF kernel. The corpora is processed in the
same way in Sects. 6.3.3.2 and 6.3.3.3, and then we call the interface of different
machine learning algorithms in sklearn. We compare the results with character-
level convolutional [97] networks which is the state-of-the-art method in text
classification. Finally we adopt the accuracy and the F1 value of text classification
as the evaluation standard.

6.3.3.5 Experiments and Results

As shown in Tables 6.6 and 6.7, the first column in the table shows the compared
training algorithms, and the first row indicates the value of λ1 . The value of
λ2 corresponding to this is 1-λ1 . The second row shows the performance of the
combined kernel in the case of different λ1 value. The rest rows represent the
performance of other machine learning methods. The values in Table 6.6 represent
the accuracy rate of Sogou corpus and the values in Table 6.7 represent the values
of F1 of Sogou corpus.
The values obtained in Tables 6.6 and 6.7 are shown by the line chart in Figs. 6.14
and 6.15 respectively from which it is easier to see the effect of the combination of
the statistical similarity matrix S and the second-order knowledge-based similarity
matrix Z 2 on the classification accuracy.
236 6 Intention Based Networking Management

Table 6.6 Accuracy rate of Sogou corpus

0 0.2 0.4 0.6 0.8 1
SVM (Combined kernel) 92.12% 94.98% 95.58% 96.43% 95.32% 93.12%
CNN (Character-level) 95.12%
SVM (Linear kernel) 75.23%
SVM (RBF kernel) 82.53%
KNN 78.62%
Bayes 85.65%

Table 6.7 F1 value of Sogou corpus

0 0.2 0.4 0.6 0.8 1
SVM (Combined kernel) 92.16% 93.68% 94.96% 95.75% 95.14% 93.36%
CNN (Character-level) 94.93%
SVM (Linear kernel) 77.32%
SVM (RBF kernel) 84.52%
KNN 78.53%
Bayes 85.12%

Fig. 6.14 Curve: accuracy rate of Sogou corpus

6.3 Sematic and Statistical Information Based Text Classification 237

Fig. 6.15 Curve: F1 value of Sogou corpus

Table 6.8 Accuracy rate of Fudan corpus

0 0.2 0.4 0.6 0.8 1
SVM (Combined kernel) 95.34% 95.75% 96.68% 96.23% 95.56% 95.14%
CNN (Character-level) 95.15%
SVM (Linear kernel) 89.41%
SVM (RBF kernel) 94.32%
KNN 90.21%
Bayes 95.59%

As shown in Fig. 6.14, the accuracy rate is lower than that using character-
level convolutional networks which is a very effective classification method in
text categorization when λ1 is 0, 0.2, and 1. However, the accuracy rate can be
maintained at a high level when λ1 is between 0.4 and 0.8. And the accuracy rate
is always higher than that using KNN, Bayes and SVM with linear kernel and
RBF kernel, proving the combination of the two is meaningful for Chinese text
classification.
The values in Table 6.8 represent the accuracy rate of Fudan corpus and the
values in Table 6.9 represent the values of F1 of Fudan corpus. The values obtained
in Tables 6.8 and 6.9 are shown by the line chart in Figs. 6.16 and 6.17, from which
we can confirm that the combination of the two kernels is meaningful for Chinese
text classification.
238 6 Intention Based Networking Management

Table 6.9 F1 value of Fudan corpus

0 0.2 0.4 0.6 0.8 1
SVM (Combined kernel) 96.12% 96.41% 96.15% 97.24% 96.58% 96.07%
CNN (Character-level) 95.37%
SVM (Linear kernel) 90.24%
SVM (RBF kernel) 94.84%
KNN 89.56%
Bayes 96.32%

Fig. 6.16 Curve: accuracy rate of Fudan corpus

6.4 Summary

In this chapter, we discuss the main challenge of intend-based networking and

introduced several algorithms. We first propose an effective model for the similarity
metrics of English sentences. Then, the SM-CHI feature selection method based
on the common method used in Chinese text classification is presented. Finally,
we proposed a novel approach which considers both the semantic and statistical
information to improve the accuracy of text classification.
References 239

Fig. 6.17 Curve: F1 value of Fudan corpus

References

1. Mikolov Tomas, Sutskever Ilya, Chen Kai, Corrado Greg S, Dean Jeff. Distributed Repre-
sentations of Words and Phrases and their Compositionality. In: Burges C. J. C., Bottou
L., Welling M., Ghahramani Z., Weinberger K. Q., eds. Advances in Neural Information
Processing Systems 26, Curran Associates, Inc. 2013 (pp. 3111–3119).
2. C. Jiang, H. Zhang, R. Yong, and H. H. Chen, “Energy-efficient non-cooperative cognitive
radio networks: Micro, meso, and macro views,” IEEE Communications Magazine, vol. 52,
no. 7, pp. 14–20, 2014.
3. Mikolov Tomas, Chen Kai, Corrado Greg, Dean Jeffrey. Efficient Estimation of Word
Representations in Vector Space. Computation and Language. 2013.
4. S. Lei, L. Zhou, Q. Peng, and H. Yao, “Openflow based spatial information network
architecture,” in International Conference on Wireless Communications & Signal Processing,
2015.
5. Collobert Ronan, Weston Jason. A unified architecture for natural language processing:
deep neural networks with multitask learning. In: International conference on machine
learning:160–167; 2008.
6. Hu Baotian, Lu Zhengdong, Li Hang, Chen Qingcai. Convolutional neural network archi-
tectures for matching natural language sentences. In: International Conference on Neural
Information Processing Systems:2042–2050; 2014.
7. Yin Wenpeng, Schütze Hinrich. Convolutional Neural Network for Paraphrase Identification.
In: Conference of the North American Chapter of the Association for Computational
Linguistics: Human Language Technologies:901–911; 2015.
8. C. Jiang, C. Yan, and K. J. R. Liu, “Graphical evolutionary game for information diffusion
over social networks,” IEEE Journal of Selected Topics in Signal Processing, vol. 8, no. 4, pp.
524–536, 2017.
240 6 Intention Based Networking Management

9. wikipedia. tf-idf. https://en.wikipedia.org/wiki/Tf%E2%80%93idf.

10. Tai Kai Sheng, Socher Richard, Manning Christopher D. Improved Semantic Representations
From Tree-Structured Long Short-Term Memory Networks. Computer Science. 2015;5(1):
36.
11. F. Xu, L. Rong, C. Zhao, H. Yao, and J. Zhang, “Congestion-aware signaling aggregation
scheme for cellular based underwater acoustic sensor network,” in IEEE International
Conference on Communication Workshop, 2015.
12. Kalchbrenner Nal, Grefenstette Edward, Blunsom Phil. A Convolutional Neural Network for
Modelling Sentences. Meeting of the association for computational linguistics. 2014;655–
665.
13. Kim Yoon. Convolutional Neural Networks for Sentence Classification. Empirical methods
in natural language processing. 2014;1746–1751.
14. He Hua, Gimpel Kevin, Lin Jimmy. Multi-Perspective Sentence Similarity Modeling with
Convolutional Neural Networks. In: Conference on Empirical Methods in Natural Language
Processing:1576–1586; 2015.
15. ——, “Multi-channel sensing and access game: Bayesian social learning with negative
network externality,” IEEE Transactions on Wireless Communications, vol. 13, no. 4, pp.
2176–2188, 2014.
16. He Hua, Lin Jimmy. Pairwise Word Interaction Modeling with Deep Neural Networks for
Semantic Similarity Measurement. In: Conference of the North American Chapter of the
Association for Computational Linguistics: Human Language Technologies:937–948; 2016.
17. H. Yao, X. Sun, Z. Zhou, L. Tang, and L. Shi, “Joint optimization of subchannel selection and
spectrum sensing time for multiband cognitive radio networks,” in International Symposium
on Communications & Information Technologies, 2010.
18. S. Lei, Z. Zheng, T. Liang, H. Yao, and Z. Jing, “Ultra-wideband channel estimation
based on Bayesian compressive sensing,” in International Symposium on Communications
& Information Technologies, 2010.
19. Socher Richard, Karpathy Andrej, Le Quoc V., Manning Chris D., Ng Andrew Y.. Grounded
compositional semantics for finding and describing images with sentences. Nlp.stanford.edu.
2013.
20. Mueller Jonas, Thyagarajan Aditya. Siamese recurrent architectures for learning sentence
similarity. In: Thirtieth AAAI Conference on Artificial Intelligence:2786–2792; 2016.
21. C. Jiang, Y. Chen, Y. H. Yang, C. Y. Wang, and K. J. R. Liu, “Dynamic Chinese restaurant
game: Theory and application to cognitive radio networks,” IEEE Transactions on Wireless
Communications, vol. 13, no. 4, pp. 1960–1973, 2014.
22. T. Liang, Z. Zhou, L. Shi, H. Yao, J. Zhang, and Y. Ye, “Laplace prior based distributed
compressive sensing,” in International 1st Conference on Communications & Networking in
China, 2010.
23. Erk Katrin. A structured vector space model for word meaning in context. In: Conference on
Empirical Methods in Natural Language Processing:897–906; 2008.
24. Z. Zhen, T. Jiang, W. S. Zhang, and H. Yao, “Analysis speech of polypus patients based on
channel parameters and fuzzy logic systems,” in Seventh International Conference on Fuzzy
Systems & Knowledge Discovery, 2010.
25. Lapata Mirella, Mitchell Jeff. Vector-based Models of Semantic Composition. 2008.
26. Mitchell J, Lapata M. Composition in distributional models of semantics. Cognitive Science.
2010;34(8):1388.
27. Turney Peter D. Domain and Function: A Dual-Space Model of Semantic Relations and
Compositions. Journal of Artificial Intelligence Research. 2013;44(4):533–585.
28. Y. H. Yang, Y. Chen, C. Jiang, C. Y. Wang, and K. J. R. Liu, “Wireless access network
selection game with negative network externality,” IEEE Transactions on Wireless Communi-
cations, vol. 12, no. 10, pp. 5048–5060, 2013.
29. H. Yao, Z. Zheng, L. He, and L. Zhang, “Optimal power allocation in joint spectrum underlay
and overlay cognitive radio networks,” in International Conference on Cognitive Radio
Oriented Wireless Networks & Communications, 2009.
References 241

30. Erk Katrin. Vector Space Models of Word Meaning and Phrase Meaning: A Survey. Language
& Linguistics Compass. 2012;6(10):635–653.
31. Clarke Daoud. A Context-theoretic Framework for Compositionality in Distributional Seman-
tics. Computational Linguistics. 2011;38(1):41–71.
32. Das Dipanjan, Smith Noah A. Paraphrase identification as probabilistic quasi-synchronous
recognition. In: Joint Conference of the Meeting of the ACL and the International Joint
Conference on Natural Language Processing of the Afnlp: Volume:468–476; 2009.
33. C. Jiang, C. Yan, and K. J. R. Liu, “Distributed adaptive networks: A graphical evolutionary
game-theoretic view,” IEEE Transactions on Signal Processing, vol. 61, no. 22, pp. 5675–
5688, 2013.
34. H. Yao, Z. Zheng, L. Zhang, L. He, T. Liang, and K. S. Kwak, “An efficient power allocation
scheme in joint spectrum overlay and underlay cognitive radio networks,” in International
Conference on Communications & Information Technologies, 2009.
35. Dolan Bill, Quirk Chris, Brockett Chris. Unsupervised construction of large paraphrase
corpora: exploiting massively parallel news sources. In: International Conference on Com-
putational Linguistics:350; 2004.
36. Nakov Preslav, Zesch Torsten, eds. Proceedings of the 8th International Workshop on
Semantic Evaluation. The Association for Computer Linguistics 2014.
37. Marelli Marco, Bentivogli Luisa, Baroni Marco, Bernardi Raffaella, Menini Stefano, Zam-
parelli Roberto. Semeval-2014 Task 1: Evaluation of compositional distributional semantic
models on full sentences through semantic relatedness and textual entailment. In: Semeval
2014: International Workshop on Semantic Evaluation:16; 2014.
38. Zhao Jiang, Zhu Tiantian, Lan Man. ECNU: One Stone Two Birds: Ensemble of Heteroge-
nous Measures for Semantic Relatedness and Textual Entailment. In: International Workshop
on Semantic Evaluation:271–277; 2014.
39. B. Li, Z. Zheng, W. Zou, K. S. Kwak, F. Wu, and H. Yao, “A nonlinear transform and its
application in the optimum receiving of ultra narrow-band,” in International Conference on
Communications & Information Technologies, 2009.
40. Bjerva Johannes, Bos Johan, Goot Rob Van Der, Nissim Malvina. The Meaning Factory:
Formal Semantics for Recognizing Textual Entailment and Determining Semantic Similarity.
In: SemEval-2014 Workshop; 2014.
41. Jimenez Sergio, Dueñas George Enrique, Baquero Julia, Gelbukh Alexander. UNAL-NLP:
Combining Soft Cardinality Features for Semantic Textual Similarity, Relatedness and
Entailment. In: Semeval; 2014.
42. Lai Alice, Hockenmaier Julia. Illinois-LH: A Denotational and Distributional Approach to
Semantics. In: International Workshop on Semantic Evaluation:329–334; 2014.
43. Bestgen Yves. CECL: a New Baseline and a Non-Compositional Approach for the Sick
Benchmark. In: International Workshop on Semantic Evaluation:160–165; 2014.
44. H. Yao, L. Zhang, Z. Ran, and Z. Zheng, “An efficient game-based competitive spectrum
offering scheme in cognitive radio networks with dynamic topology,” in IEEE International
Conference on Communication Technology, 2008.
45. Proisl Thomas, Evert Stefan, Greiner Paul, Kabashi Besim. SemantiKLUE: Robust Semantic
Similarity at Multiple Levels Using Maximum Weight Matching. In: International Workshop
on Semantic Evaluation:532–540; 2014.
46. Rus Vasile, Mccarthy Philip M., Lintean Mihai C., Mcnamara Danielle S., Graesser Arthur C.
Paraphrase Identification with Lexico-Syntactic Graph Subsumption. In: International Florida
Artificial Intelligence Research Society Conference, May 15–17, 2008, Coconut Grove,
Florida, USA:201–206; 2008.
47. Blacoe William, Lapata Mirella. A comparison of vector-based representations for semantic
composition. In: Joint Conference on Empirical Methods in Natural Language Processing and
Computational Natural Language Learning:546–556; 2012.
48. Z. Ran, L. Zhang, Y. Fei, H. Yao, and Z. Zheng, Balanced clustering multi-hop routing
algorithm for LEACH protocol in wireless sensor networks, 2008.
242 6 Intention Based Networking Management

49. Fernando Samuel, Stevenson Mark. A Semantic Similarity Approach to Paraphrase Detection.
Computational Linguistics UK Annual Research Colloquium. 2008.
50. Yao Haipeng, Liu Chong, Zhang Peiying, Wang Luyao. A feature selection method based on
synonym merging in text classification system. Eurasip Journal on Wireless Communications
& Networking. 2017;2017(1):166.
51. C. Jiang, C. Yan, G. Yang, and K. J. R. Liu, “Joint spectrum sensing and access evolutionary
game in cognitive radio networks,” IEEE Transactions on Wireless Communications, vol. 12,
no. 5, pp. 2470–2483, 2013.
52. Hwee Tou Ng, Wei Boon Goh, and Kok Leong Low. Feature selection, perceptron learning,
and a usability case study for text categorization. In International ACM SIGIR Conference on
Research and Development in Information Retrieval, pages 67–73. ACM, 1997.
53. Yiming Yang and Jan O Pedersen. A comparative study on feature selection in text
categorization. In Fourteenth International Conference on Machine Learning, pages 412–
420. Morgan Kaufmann Publishers Inc, 1997.
54. Gui Chuan Feng and Shubin Cai. An improved feature extraction algorithm based on chi and
mi. 2015.
55. Yan Tang and Ting Xiao. An improved χ 2 (chi) statistics method for text feature selection.
In International Conference on Computational Intelligence and Software Engineering, pages
1–4. IEEE, 2009.
56. Thorsten Joachims. Text categorization with Support Vector Machines: Learning with many
relevant features. Springer Berlin Heidelberg, 1998.
57. H. P. Yao, L. Y. Zhang, Z. Zhou, and X. U. Fang-Min, “A handoff algorithm based on grey
prediction model for bluetooth network,” Radio Engineering of China, 2008.
58. C. Jiang, Y. Chen, K. J. R. Liu, and Y. Ren, “Renewal-theoretical dynamic spectrum access in
cognitive radio networks with unknown primary behavior,” IEEE Journal on Selected Areas
in Communications, vol. 31, no. 3, pp. 406–416, 2013.
59. Ted Dunning. Accurate methods for the statistics of surprise and coincidence. Linguistics-74
Computational Dirk Geeraerts Stefan Grondelaers and, 19(1):61–74, 1993.
60. J. Tian and W. Zhao. Words similarity algorithm based on tongyici cilin in semantic web
adaptive learning system. Journal of Jilin University, 28(06):602–608, 2010.
61. H. Yao, Z. Zhang, and Y. Liu, “Research on the embedded sim technology in internet of
things,” Information & Communications Technologies, 2012.
62. Sijun Qin, Jia Song, Pengzhou Zhang, and Yue Tan. Feature selection for text classification
based on part of speech filter and synonym merge. In International Conference on Fuzzy
Systems and Knowledge Discovery, pages 681–685, 2015.
63. H. Yao, Z. Yang, H. Jiang, and L. Ma, “A scheme of ad-hoc-based d2d communication in
cellular networks.” Adhoc & Sensor Wireless Networks, vol. 32, 2016.
64. Susan Dumais, John Platt, David Heckerman, and Mehran Sahami. Inductive learning algo-
rithms and representations for text categorization. In Proceedings of the seventh international
conference on Information and knowledge management, pages 148–155. ACM, 1998.
65. Quoc V. Le and Tomas Mikolov. Distributed representations of sentences and documents.
4:II–1188, 2014.
66. Jeffrey Pennington, Richard Socher, and Christopher Manning. Glove: Global vectors for
word representation. In Conference on Empirical Methods in Natural Language Processing,
pages 1532–1543, 2014.
67. Bernhard E Boser, Isabelle M Guyon, and Vladimir N Vapnik. A training algorithm for
optimal margin classifiers. In Proceedings of the fifth annual workshop on Computational
learning theory, pages 144–152. ACM, 1992.
68. Stephan R Sain. The nature of statistical learning theory. Technometrics, 38(4):409–409,
1996.
69. Thorsten Joachims. Text categorization with support vector machines: Learning with many
relevant features. In European conference on machine learning, pages 137–142. Springer,
1998.
References 243

70. Shun-ichi Amari and Si Wu. Improving support vector machine classifiers by modifying
kernel functions. Neural Networks, 12(6):783–789, 1999.
71. Jamal Abdul Nasir, Asim Karim, George Tsatsaronis, and Iraklis Varlamis. A knowledge-
based semantic kernel for text classification. In International Symposium on String Processing
and Information Retrieval, pages 261–266. Springer, 2011.
72. Berna Altınel, Banu Diri, and Murat Can Ganiz. A novel semantic smoothing kernel for text
classification with class-based weighting. Knowledge-Based Systems, 89:265–277, 2015.
73. George Siolas and Florence d’Alché Buc. Support vector machines based on a semantic kernel
for text categorization. In Neural Networks, 2000. IJCNN 2000, Proceedings of the IEEE-
INNS-ENNS International Joint Conference on, volume 5, pages 205–209. IEEE, 2000.
74. Dimitrios Mavroeidis, George Tsatsaronis, Michalis Vazirgiannis, Martin Theobald, and
Gerhard Weikum. Word sense disambiguation for exploiting hierarchical thesauri in text
classification. In European Conference on Principles of Data Mining and Knowledge
Discovery, pages 181–192. Springer, 2005.
75. C Fellbaum and G Miller. WordNet:An Electronic Lexical Database. MIT Press, 1998.
76. Yan-Lan Zhu, Jin Min, Ya-qian Zhou, Xuan-jing Huang, and Li-De Wu. Semantic orientation
computing based on hownet. Journal of Chinese Information Processing, 20(1):14–20, 2006.
77. Pei-Ying Zhang. A hownet-based semantic relatedness kernel for text classification. Indone-
sian Journal of Electrical Engineering and Computer Science, 11(4):1909–1915, 2013.
78. Jiaju Mei. Tongyi ci cilin. Shangai cishu chubanshe, 1985.
79. Nicholas E. Evangelopoulos. Latent semantic analysis. Wiley interdisciplinary reviews.
Cognitive science, 4(6):683, 2013.
80. Berna Altlnel, Murat Can Ganiz, and Banu Diri. A novel higher-order semantic kernel for
text classification. In Electronics, Computer and Computation (ICECCO), 2013 International
Conference on, pages 216–219. IEEE, 2013.
81. Berna Altinel, Murat Can Ganiz, and Banu Diri. A semantic kernel for text classification based
on iterative higher-order relations between words and documents. In International Conference
on Artificial Intelligence and Soft Computing, pages 505–517. Springer, 2014.
82. Berna Altinel, Murat Can Ganiz, and Banu Diri. A simple semantic kernel approach for svm
using higher-order paths. In Innovations in Intelligent Systems and Applications (INISTA)
Proceedings, 2014 IEEE International Symposium on, pages 431–435. IEEE, 2014.
83. Berna Altınel, Murat Can Ganiz, and Banu Diri. A corpus-based semantic kernel for text
classification by using meaning values of terms. Engineering Applications of Artificial
Intelligence, 43:54–66, 2015.
84. H. Yao, Y. Liu, and C. Fang, “An abnormal network traffic detection algorithm based on
big data analysis.” International Journal of Computers, Communications & Control, vol. 11,
no. 4, 2016.
85. Karen Sparck Jones. A statistical interpretation of term specificity and its application in
retrieval. Journal of documentation, 28(1):11–21, 1972.
86. Gerard Salton and Christopher Buckley. Term-weighting approaches in automatic text
retrieval. Information processing & management, 24(5):513–523, 1988.
87. Youngjoong Ko and Jungyun Seo. Automatic text categorization by unsupervised learning. In
Proceedings of the 18th conference on Computational linguistics-Volume 1, pages 453–459.
Association for Computational Linguistics, 2000.
88. Verayuth Lertnattee and Thanaruk Theeramunkong. Analysis of inverse class frequency in
centroid-based text classification. In Communications and Information Technology, 2004.
ISCIT 2004. IEEE International Symposium on, volume 2, pages 1171–1176. IEEE, 2004.
89. Göksel Biricik, Banu Diri, Ahmet Co, et al. A new method for attribute extraction with appli-
cation on text classification. In Soft Computing, Computing with Words and Perceptions in
System Analysis, Decision and Control, 2009. ICSCCW 2009. Fifth International Conference
on, pages 1–4. IEEE, 2009.
90. GÖKSEL BİRİCİK, Banu Diri, and AHMET COŞKUN SÖNMEZ. Abstract feature extrac-
tion for text classification. Turkish Journal of Electrical Engineering & Computer Sciences,
20(Sup. 1):1137–1159, 2012.
244 6 Intention Based Networking Management

91. Simon Parsons. Introduction to machine learning by ethem alpaydin, MIT press, 0-262-
01211-1, 400 pp., $50.00/£ 32.95, 2005.
92. Nello Cristianini, John Shawe-Taylor, and Huma Lodhi. Latent semantic kernels. Journal of
Intelligent Information Systems, 18(2-3):127–152, 2002.
93. Edward Loper and Steven Bird. Nltk: the natural language toolkit. In Acl-02 Workshop
on Effective TOOLS and Methodologies for Teaching Natural Language Processing and
Computational Linguistics, pages 63–70, 2002.
94. Fabian Pedregosa, Alexandre Gramfort, Vincent Michel, Bertrand Thirion, Olivier Grisel,
Mathieu Blondel, Peter Prettenhofer, Ron Weiss, Vincent Dubourg, and Jake Vander-
plas. Scikit-learn: Machine learning in python. Journal of Machine Learning Research,
12(10):2825–2830, 2011.
95. Mohamed El Kourdi, Amine Bensaid, and Tajje Eddine Rachidi. Automatic arabic document
categorization based on the naïve bayes algorithm. In The Workshop on Computational
Approaches To Arabic Script-Based Languages, pages 51–58, 2004.
96. Mostafa M Syiam, Zaki T Fayed, and Mena B Habib. An intelligent system for Arabic
text categorization. International Journal of Intelligent Computing and Information Sciences,
6(1):1–19, 2006.
97. Xiang Zhang, Junbo Zhao, and Yann Lecun. Character-level convolutional networks for text
classification. pages 649–657, 2015.
Chapter 7
Conclusions and Future Challenges

7.1 Conclusions

This book mainly discusses the architectures of intelligent networks as well as the
possible techniques and challenges. As shown in Fig. 7.1, we first introduced the
concept of NetworkAI, which is a novel paradigm that applying machine learning
to automatically control a network. NetworkAI employs reinforcement learning
and incorporates network monitoring technologies such as the in-band network
telemetry to dynamically generate control policies and produces a near-optimal
decision. We employ the SDN and INT to implement a network state upload link
and a decision download link to accomplish a closed-loop control of a network and
build a centralized intelligent agent aiming at learning the policy by interaction with
a whole network.
Then, we discuss the possible machine learning methods for network awareness.
With the rapid development of compelling application scenarios of the networks,
such as 4K/8K, IoT, it becomes substantially important to strengthen the manage-
ment of data traffic in networks. As a critical part of massive data analysis, traffic
awareness plays an important role in ensuring network security and defending traffic
attacks. Moreover, the classification of different traffic can help to improve their
work efficiency and quality of service (QoS).
Furthermore, we discuss how machine learning can achieve network automati-
cally control. One of the most critical issues in the network is finding a near-optimal
control strategy. Examples include routing decision, load balancing, QoS-enable
load scheduling, and so on. However, the majority solutions of these problems are
largely relying on a manual process.
Therefore, to address this issue, in Chap. 4, we apply several artificial intelligence
approaches for self-learning control strategies in networks. In addition, resource
management problems are ubiquitous in the networking field, such as job schedul-
ing, bitrate adaptation in video streaming and virtual machine placement in cloud

© Springer Nature Switzerland AG 2019 245

H. Yao et al., Developing Networks using Artificial Intelligence,
Wireless Networks, https://doi.org/10.1007/978-3-030-15028-0_7
246 7 Conclusions and Future Challenges

natural
language
understa
nding
natural
language Intent-based networking
Users

Resource management
Traffic
Traffic Engineering
QoS\QoE
VNE
Traffic
QoS\QoE
Network State Machine learning Routing

Network Analytics platform

Network Network
Awareness SDN Controller Control

Fig. 7.1 NetworkAI conception

computing. In Chap. 5, we focus on how machine learning can optimal resource

management decision.
Finally, we discuss the intent-based networking which allows a user or admin-
istrator to send a simple request—using natural language—to plan, design and
implement/operate the physical network, so that it can reduce network complexity
and improves automation by eliminating manual configurations. For example, an IT
administrator can request improved voice quality for its voice-over-IP application,
and the network can respond. For intent-based networking, the translation and
validation system take a higher-level business policy (what) as input from end users
and converts it to the necessary network configuration (how) by natural language
understanding technology. To achieve intent-based networking, in Chap. 6, we focus
7.2 Future Challenges 247

on how machine learning can technology can be used in the natural language
understanding in translation and validation system.

7.2 Future Challenges

7.2.1 New Hardware Architecture

Every innovation of the upper-level services is based on the extra improvement

in the underlying hardware, for example, the Central Processing Unit (CPU) for
general purpose computer, Digital Signal Processing (DSP) for the communication
system, and Graphics Processing Unit for image processing. Similarly, for meeting
the AI-driven networking age, the requirement for the specific AI networking
processor is obviously.
Nowadays, networks generate millions of different types of flows every millisec-
ond. Running AI agent in such massive volumes data is extremely challenging. The
currently routers’ computing power is far from being satisfied the requirement of
AI&ML deployment. Especially, machine learning is a high computational intensive
process. Recently, as the highly parallel, multi-core, multi-threaded processor, the
GPU/TPU chip become the cornerstone for the age of AI. And some research works
have shown that the GPU can improve the packet processing capability. However,
due to the high-speed mass data processing (more than 10 Gb/s) and the demanding
respond delay (less than 1 ms) in the future network, there is still a big gap between
universal AI processing chip and actual deployment in the networking field.

7.2.2 Advancing Software System

Currently, the network data is becoming a big data challenges, e.g. threefold
increase in total IP Traffic, >60% increase in devices and connections, telemetry
data streamed in near real-time. Meanwhile, the geo-distributed characteristics of
networking further impose more difficulties for the widespread deployment of
network data analytic platform. For example, how to aggregates data like logs,
metrics, and network telemetry; how to scales up to consume millions of flows
per second; how to efficiently share knowledge among distributed network nodes.
The currently end-to-end solution, which combines multiple technologies such
as Apache Spark and Hadoop MapReduce, can be extremely complex and time-
consuming. Therefore, a powerful scalable big data&AI analytic platform for
networks and services is needed.
In addition, the software library for networking machine learning tasks is another
enabler for AI-based networking. Machine learning framework offers a high-level
programming interface for designing, training and validating machine learning.
248 7 Conclusions and Future Challenges

However, current machine learning frameworks, such as Tensor-Flow, Caffe, and

Theano, are designed for general-purpose, which is too overhead for networking
domain. It needs further optimize to satisfy the requirement of the networking field,
such as processing speed, low complexity, and light-weight.

7.2.3 Self-Adaptive Network Protocol

In networks, for exchanging various messages, the network protocols define rules,
procedures, and formats for communication among network nodes. The current
network protocols largely rely on the human being defined. To improve network
flexibility and efficiency, there is still enough room for redesigning the network
protocol to accelerate the efficiency of exchanging messages mechanism. Rather
than matching patterns in a huge corpus of text, recently OpenAI released initial
result that they trained the AI agents to invent a new language which is grounded
and compositional. The multiple agents can automatically learn the communication
protocol to coordinate their act enabled by machine learning. While these results are
still far from the success of network protocol self-learning, it gives us a possibility
of future network protocol evolution tendency.

7.2.4 Promoting Machine Learning Algorithms

While the machine learning algorithms are like a hundred flowers in bloom,
current machine learning algorithms are driven by the existing business applica-
tion, such as computer vision and natural language understanding. For example,
convolutional neural networks are fascinating and powerful in image and audio
recognition and even achieve superhuman performance in many tasks. However,
the network presents a totally different theoretical mathematic model compared to
the vision/NLP fields. The convolutional layers or recurrent layer may not work
effectively in the networking domain. In addition, the network has massively more
data and demanding response time which poses great challenges for machine learn-
ing deployment. Therefore, the demanding requirement and specific characteristics
of networking require both adapting existing algorithms and developing new ones.
The networking as a new application for machine learning will push forward of both
machine learning and networking domain to a new stage.