JAIN (DEEMED-TO-BE UNIVERSITY)

BENGALURU, KARNATAKA

P2: Teaching-Learning and Evaluation Plan

PROGRAMME: MASTER OF COMPUTER APPLICATIONS

Teaching - Learning & Evaluation Plan

Course Information:
Course Code: 23MCAAI302 Course Title: DEEP LEARNING
Credits Units: 3 Total Contact Hours: 45 L-T-P: 3-0-0
IA: ESE Weightage- 50:50 Pass Marks (IA and ESE): NA:18 Aggregate Pass Marks: 50%

ESE Question Paper Marks: 50 Special Examination Fees: NA Pre-requisite (if any): NA

Course Facilitator: Dr. Prerna Mahajan, Professor, School of CS & IT

Programme Outcomes (POs) and Programme Specific Outcomes (PSOs)

At the end of the Programme, students will be able to

PROGRAMME OUTCOMES
PO 1
Computational Knowledge: Apply mathematical foundation, computing and domain
knowledge for the conceptualization of computing models from defined problems.
PO 2
Problem Analysis: Analyze complex computing problems using fundamentals of computer
science and application domains.
PO 3 Design &Develop Solutions: Transform complex business scenarios and contemporary issues
into problems and propose integrated solutions using emerging technologies.
 PO 4
Conduct Investigations of Complex Computing Problems: conduct experiments, interpret
data and provide well informed conclusions.
PO 5
Modern Tool Usage: Select modern computing tools, skills and techniques necessary for
providing innovative software solutions.
PO 6
Innovation and Entrepreneurship: Create ideas for opportunities, entrepreneurship vision
and use of wealth for the betterment of the individual and society.
Societal & Environmental Concern: Recognize economical, environmental, social, health,
PO 7 legal, ethical issues involved in the usage of computingtechnologies and other consequential
responsibilities relevant to professional practice.
PO 8
Professional Ethics: Employ professional ethics and cyber regulations in a global economic
environment.
PO 9
Individual & Team Work: Function effectively as a member or leader in diverse teams in
multidisciplinary environment.
PO 10
Communication Efficacy: Communicate effectively with the computing community as well
as society by being able to comprehend effective documentations and presentations.
PO11
Project Management: Demonstrate managerial skillswith computing knowledge to manage
projects in multidisciplinary environments.
PO12
Life-long Learning: Develop the ability and acquire the skills required to engage in
continuous learning as a Computing professional.

PROGRAMME SPECIFIC OUTCOMES

PSO1
Acquire proficiency in developing and implementing effective solutions using new age IT
platforms and emerging technologies.
Develop professional and ethical attitude, effective communication skills, team work skills
PSO2 and multidisciplinary approach to relate computer solutions to broader social context there by
enhancing entrepreneurship skills.

Course Objectives:
1 COB1 To teach the Numerical Computations behind neural networks.
2 COB2 To provide the knowledge about the different layers of
Convolutional Neural Networks.
3 COB3 To explain the underlying mathematical operations of every
layer in a generic Neural Network.
4 COB4 To execute the processing of Auto Encoders in real-time
 problems.
5 COB5 To develop the knowledge of Reinforcement Learning Models
and its importance.
Course Outcomes:
At the end of the course, students will be able to:

Sl. Course Description Bloom’s Taxonomy

No Outcome Level
.
1. CO 1 Describe the feed-forward and deep networks. UNDERSTANDING
(LEVEL 2)
2. CO 2 Design single and multi-layer feed-forward deep APPLYING
networks and tune various hyper-parameters. (LEVEL 3)
3. CO 3 Illustrate the working of a generic Neural Network. ANALYZING
(LEVEL4)
4. CO 4 Develop an Auto Encoder for a real-time problem. CREATING
(LEVEL 6)
5. CO 5 Illustrate the process of Reinforcement Learning. EVALUATING
(LEVEL5)

CO-PO/PSO Mapping:
PO PO PO PO PO PO PO PO PO PO PO PO PS PSO2
1 2 3 4 5 6 7 8 9 10 11 12 O1

CO1 3 2 - - - - 1 - - - - - 3 -
CO2 3 2 - 3 - - 1 - - - - - 3 3
CO3 - 3 3 3 1 - 1 - 1 - 2 - 3 -
CO4 - 3 3 3 3 - 2 - - - 2 2 3 -
CO5 - 2 2 2 3 - 2 1 - - - 2 3 -
Course Contents:
SYLLABUS
 MODULE CONTENTS Assessments and CO PO Mapping
NO. Activity

Introduction to Practical Deep Networks:

Assignment CO1 PO1
MODULE – Probability and Information Theory – Numerical
1 Computation - Deep feed-forward Networks –
Regularization for Deep Learning – Optimization for
Training Deep Models.

Mathematical Foundation for Deep Learning:

Scalars (0D Tensor), Vectors (1D Tensor), Matrices Assignment and CO2 PO2
MODULE – (2D Tensors), Higher Dimension Tensors, Assessment on
2 Manipulating Tensors, Tensor Operations, Gradient LinkedIn
Based Optimization: Derivative of a Tensor Certification Course
operation, Stochastic Gradient Descent, Chaining
Derivative: Backpropagation Algorithm.

Convolutional Neural Networks:

Case Study on CO3 PO3
MODULE – Feed-forward neural networks, deep networks, various
3 regularizing a deep network, model exploration, and
hyper-parameter tuning. Tensor
Introduction to convolution neural networks: Operations
stacking, striding and pooling, applications like
image, and text classification.
 Recurrent Neural Network:
MODULE Group / Individual
–4 Structure of RNN, Component of Presentation on Impact of
RNN, Backpropagation in RNN, Business Implica- tions of CO3, PO2, PO4. PO7
Vanishing Gradient and Exploding Cognitive Computing. CO4
Gradient.
Long Short Term Memory (LSTM):
Limitation of RNN, structure of LSTM,
Components of LSTM, Backpropagation
through time.
Reinforcement Learning:
MODULE Case Study on a Real time CO4 PO4, PO5
Learning scenario - Markov decision process prob- lem that uses
–5
model - Policy - Planning al- gorithms - Reinforcement Learning
Learning algorithms: Stochastic approximation
- TD(0) algo- rithm - Q-learning algorithm -
SARSA - TD(λ) algorithm- Large state space.

TEXT BOOKS:
1. Ian Goodfellow, Deep Learning, 1st Edition, MIT Press, 2116.
2. Reinforcement Learning: An Introduction, Richard S. Sutton & Andrew G. Barto, 3rd edition, 2019.
3. Deep Learning with Python, Second Edition, Francois Chollet, 2nd edition, 2018.

REFERENCES:
1. Jeff Heaton, Deep Learning and Neural Networks, Heaton Research Inc, 2015.
2. Mindy L Hall, Deep Learning, VDM Verlag, 2011
 E – RESOURCES:
● [Link]
● [Link]
● [Link]

ONLINE COURSES:
● [Link]
● [Link]

Program Outcome / Course Outcome Mapping CO-PO Mapping (3-strong, 2-

Good, 1-Weak)

PO PO PO PO PO PO PO PO PO PO PO PO PS PSO2
1 2 3 4 5 6 7 8 9 10 11 12 O1

CO1 3 2 - - - - 1 - - - - - 3 -
CO2 3 2 - 3 - - 1 - - - - - 3 3
CO3 - 3 3 3 1 - 1 - 1 - 2 - 3 -
CO4 - 3 3 3 3 - 2 - - - 2 2 3 -
CO5 - 2 2 2 3 - 2 1 - - - 2 3 -

Text Books and Reference Books

TB1. Patterson, Josh, and Adam Gibson. “Deep learning: A practitioner's approach.”
O'Reilly Media, Inc.", 2017.
TB2. Moolayil, Jojo, JojoMoolayil, and Suresh John. “Learn Keras for Deep Neural
Networks.” Birmingham: Apress, 2019.
REFERENCE BOOKS:
R1. Goodfellow, Ian, YoshuaBengio, and Aaron Courville. “Deep learning.” MIT press,
2016.
R2. Osinga, Douwe. “Deep learning cookbook: Practical recipes to get started quickly.”
O'Reilly Media, Inc.", 2018.
R3. Gulli, Antonio, and Sujit Pal. “Deep learning with Keras.” Packt Publishing Ltd, 2017.

Session-Wise Plan:

Mode
Sessi Pedagogy/ of
Topic Readings and References CO
on Activity Planned Deliver
y
1-3 Probability and Information TB1,R1 Presentation on 1 In class
Theory various discussi
on and
 Regularization PPT
Techniques

TB1,R1
Numerical Computation-
4-5 1
Probability concepts

Deep feed-forward Networks- TB1,R1

review of activation functions,
6-7 1
architecture, feed forward
loop
Regularization for Deep TB1,R1
Learning- epochs/dropout
8-10 1
layer, other regularisation
methods
Optimization for Training
11-12 Deep Models-Adam/Adagrid TB1,R1 1
etc
Activity-1
TB1,R1
13 Tensors- Overview 2

TB1,R1
Scalars (0D Tensor),
14 Vectors,Matrices, Higher 2
Dimension Tensor
TB1,R1
15 Manipulating Tensors 2

TB1,R1 LinkedIn
In class
Certification
Tensor Operations, Derivative discussi
16 Course and 2
of a Tensor operation on and
Report PPT
submission
TB1,R1
17 Gradient Based Optimization: 2

TB1,R1
18 Stochastic Gradient Descent 2

TB1,R1
Chaining Derivative:
19 2
Backpropagation Algorithm

Activity-2
TB1,R1 Explore the In class
Feed-forward neural concepts of discussi
20 networks, structural convolution 3 on and
Limitations neural networks PPT
 TB1,R1
Deep networks: features in
21 3
respect of image data

TB1,R1
Regularizing a deepnetwork,
22 Model exploration, and 3
hyper-parameter tuning.
TB1,R1
Introduction to convolution
21 3
neural networks

TB1,R1
22 Stacking, striding andpooling 3

TB1,R1
Applications like image, and
23 3
text classification

Activity-3
TB1,R1 Group In class
Recurrent Neural Presentation on discussi
24 Network:Overview , context CNN and RNN 4 on and
for text processing PPT
TB1,R1
25 Structure and features of RNN 4

TB1,R1
26 Components of RNN 4

TB1,R1
27 Backpropagation inRNN 4

TB1,R1
Vanishing Gradient and
28 4
Exploding Gradient

TB1,R1
Long Short Term Memory
29 (LSTM) 4

TB1,R1
Limitation of RNN, structure
30 4
of LSTM

TB1,R1
31 Components of LSTM 4
 TB1,R1
32 Backpropagation through time 4

TB2,R1
33 Learning Scenario-Markov 5
Decision Process Model

TB2,R1
Markov Decision Process
34 5
Model(ctd)

TB2,R1
Quality Policy, State, Action
35 5
etc.

TB2,R1
Quality Policy, State, Action
36 5
etc.(ctd)

TB2,R1
37 Planning Algorithtms LinkedIn 5
In class
Certification
discussi
Course and
TB2,R1 on and
Report PPT
38 Stochasttic Approximation submission 5

TB2,R1
39 TD(0) Algorithm 5

TB2,R1
40 Q-Learning Algorithm 5

TB2,R1
41 SARSA-TDAlgorithm 5

TB2,R1
42 SARSA-TDAlgorithm(ctd) 5

TB2,R1
Large state space and
43 5
considerations

Revision and Doubt Clearing

44
Session
 Revision and Doubt Clearing
45
Session

Assessment Scheme: CA:ESE - 50:50

Sl. Assessment Formative/ Frequency/Stages Weightage CO
No. Instrument Summative (%)
1. Activity-1 1 10 CO1,CO2,
2. Activity-2 Formative 1 10 CO3

3. Activity-3 1 10 CO4,CO5

3. Class Participation Formative - 5 CO1-CO5

4. Periodic Test Formative 3 15 CO1-CO5
5. End Semester Examination Summative 1 50 CO1-CO5
Total 100

Assessment Sheet with Rubrics for Grading & Evaluation

Activity-1: presentation

Presentation Report on time

Sr. Student Presentation Content Skills submission Total Conversion
No. USN No. Name 10
3 Marks 4 Marks 3 Marks Marks 10 Marks

Activity-2: Certification

Learning Report on time

Sr. Student Certification Outcome Report submission Total Conversion
No. USN No. Name 10
5 Marks 3 Marks 2 Marks Marks 10 Marks

Activity-3: Case Study/Model Computation and Evaluation

Model Model Report on time

Sr. Student Development Evaluation submission Total Conversion
No. USN No. Name 10
5 Marks 2 Marks 3 Marks Marks 10 Marks

WEB LINKS
1. [Link]
2. [Link]
3. [Link]