6.207/14.

15: Networks
Lectures 2 & 3: Graphs, Measures and Metrics

1
 Networks: Lectures 2 & 3 Introduction

Outline

◦ Network representation
− Graphs. Definitions. Notations.

◦ Graph properties, measurements and metrics

− Diameter, Average Path Length, Degree Distributions
− Clustering Coefficient, Centrality

Reading:
− Newman, Chapter 6 (skip Sections 6.8 and 6.12-6.14).
− Newman, Chapter 7, Sections 7.1, 7.6, and 7.7.

2
 Networks: Lectures 2 & 3 Introduction

Network Study

◦ Historical study of networks:

− Mathematical graph theory: One of the pillars of discrete mathematics
− Started with Euler’s 1735 solution of the Königsberg bridge problem.
− Can you cross each bridge exactly once in a walk?

Figure: Islands in Königsberg, Prussia connected via Seven Bridges.

− Networks also studied extensively in sociology.

− Typical studies involve circulation of questionnaires/small networks.
Images © Bogdan Giuşcă, Mark Foskey, and Xiong on Wikipedia. All rights reserved. This content is excluded from our Creative Commons license. For more
information, see https://ocw.mit.edu/help/faq-fair-use/ 3
 Networks: Lectures 2 & 3 Introduction

Network Study

◦ Recent years witnessed a substantial change in network research.

− From analysis of single small graphs (10-100 nodes) to statistical
properties of large scale networks (million-billion nodes).
− Motivated by availability of computers and computer networks that
allow us to gather and analyze large scale data.

◦ New Analytical Approach:

− Find statistical properties that characterize the structure of these
networks and ways to measure them
− Create models of networks (structure and dynamics)
− Predict behavior of networks on the basis of measured structural
properties and models

4
 Networks: Lectures 2 & 3 Graphs

Graphs: Network Representation

◦ A network G (also called a graph) is a set of nodes N = {1, . . . , n} joined

by edges (or links).
◦ We will be mostly focusing on simple graphs: a graph with no self-edges or
multi edges.
◦ A network is typically represented by its adjacency matrix which is an n × n
matrix A = [Aij ]i ,j ∈N , where
(
1 if there is an edge from j to i,
Aij =
0 otherwise.

− The edge weight Aij can take on non-binary values (even negative
values), representing the intensity of the interaction, in which case we
refer to G as a weighted graph.
− For simple graphs, diagonal elements are zero.
◦ We refer to a graph as a directed graph (or digraph) if Aij 6= Aji and an
undirected graph if Aij = Aji for all i, j ∈ N.
5
 Networks: Lectures 2 & 3 Graphs

Graphs: Network Representation

1
 
0 0 1
Example 1:  1 0 0  ⇒
0 1 0
2 3

1 1
 
0 1 1
Example 2:  1 0 1  ⇒ ⇒
1 1 0
2 3 2 3

6
 Networks: Lectures 2 & 3 Graphs

Graphs: Network Representation

◦ Another representation of a graph is given by G = (N, E ), where

E = {1, . . . , m } is the set of edges in the network.
− For directed graphs: E is the set of “directed” edges, (i, j ) ∈ E .
− For undirected graphs: E is the set of “undirected” edges,{i, j } ∈ E .

◦ In Example 1, Ed = {(1, 2), (2, 3), (3, 1)}


◦ In Example 2, Eu = {1, 2}, {1, 3}, {2, 3}

◦ When are directed/undirected graphs applicable?

− Citation networks: directed
− Friendship networks: undirected

7
 Networks: Lectures 2 & 3 Graphs Properties

Walks, Paths, and Cycles

◦ We consider “sequences of edges” to capture indirect interactions.

◦ For an undirected graph G :
− A walk is a sequence of edges {i1 , i2 }, {i2 , i3 }, . . . , {iK −1 , iK }.
− A path between nodes i and j is a sequence of edges
{i1 , i2 }, {i2 , i3 }, . . . , {iK −1 , iK } such that i1 = i and iK = j, and each
node in the sequence i1 , . . . , iK is distinct.
− A cycle is a path with a final edge to the initial node.
− A geodesic between nodes i and j is a “shortest path” (i.e., with
minimum number of edges) between these nodes.
◦ A path is a walk where there are no repeated nodes.
◦ The length of a walk (or a path) is the number of edges on that walk (or
path).
◦ For directed graphs, the same definitions hold with directed edges (in which
case we say “a path from node i to node j”).

8
 Networks: Lectures 2 & 3 Graphs Properties

Walks, Paths, and Cycles

i j i j i j i j

walk path between i and j cycle shortest path

◦ Note: For simple graphs, we can calculate the number of walks of

(r )
given length r between nodes i and j, Nij , using the adjacency
matrix:
n
∑ Aik Akj = [A2 ]ij .
(2)
Nij =
k =1
(r )
◦ Similarly, Nij = [Ar ]ij .

9
 Networks: Lectures 2 & 3 Graphs Properties

Connectivity and Components

◦ An undirected graph is connected if every two nodes in the network

are connected by some path in the network.
◦ Components of a graph (or network) are the distinct maximally
connected subgraphs.
◦ A directed graph is
− connected if the underlying undirected graph is connected (i.e.,
ignoring the directions of edges).
− strongly connected if each node can reach every other node by a
“directed path”.

2 3

Figure: A directed graph that is connected but not strongly connected

10
 Networks: Lectures 2 & 3 Graphs Properties

Special Graphs

◦ Hypergraphs: Graphs in which edges join more than two nodes (such
edges are called hyper edges).
− A social network representing families in a village.
− Any network in which nodes connected by common membership of
groups (also called affiliation networks).

◦ Bipartite graphs: Graphs in which nodes decompose into two groups

such that there are edges only between these groups.
− Hypergraphs can be represented as a bipartite graph.

◦ A tree is a connected (undirected) graph with no cycles.

− In a tree, there is a unique path between any two nodes.
− A connected graph is a tree if and only if it has n − 1 edges.

11
 Networks: Lectures 2 & 3 Graphs Properties

Special Graphs

Complete graph Ring Star

Bipartite graph

Tree actors movies

12
 Networks: Lectures 2 & 3 Graphs Properties

Neighborhood and Degree of a Node

◦ The neighborhood of node i is the set of nodes that i is connected to.

◦ For undirected graphs:
− The degree of node i is the number of edges connected to i (i.e.,
cardinality of his neighborhood).
− The degree of node i, ki , can be written in terms of the adjacency
matrix as
n
ki = ∑ Aij .
j =1

− An important relation that is used extensively relates number of edges

to sum of degrees in the graph:
n
2m = ∑ ki .
i =1

1
− Average node degree is given by c = n ∑ni=1 ki , or equivalently c = 2nm .
13
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Neighborhood and Degree of a Node

◦ The maximum number of edges in a simple graph is (n2) = n(n2−1) .

◦ We will sometimes consider networks in which all nodes have the
same degree. Such networks are called regular networks.
◦ For directed graphs:
− Node i’s in-degree is ∑nj =1 Aij (number of incoming edges).
− Node i’s out-degree is ∑nj =1 Aji (number of outgoing edges).

1 2

4 3

Figure: Node 1 has in-degree 1 and out-degree 2.

14
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Konigsberg
¨ bridge problem

◦ Can you cross each bridge exactly once in a walk?

Figure: Islands in Konigsberg,

¨ Prussia connected via Seven Bridges.

◦ Euler’s insights:
− Graph matters, not physical properties of the island, size, etc.
− Suppose such a walk existed: starts at “node” u, ends at v . Then
− Nodes u and v can have odd degrees
− All other nodes must have even degrees
− What about Konigsberg
¨ bridge graph?
Images © Bogdan Giuşcă, Mark Foskey, and Xiong on Wikipedia. All rights reserved. This content is excluded from our Creative Commons license. For mor
information, see https://ocw.mit.edu/help/faq-fair-use/ 15
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Degree Distributions

◦ The degree distribution, P (d ), of a network is a description of relative

frequencies of nodes that have different degrees d.
− For a given graph: P (d ) is a histogram, i.e., P (d ) is the fraction of
nodes with degree d.
− For a random graph model: P (d ) is a probability distribution.
◦ Two interesting degree distributions:
− P (d ) ≤ c e −αd , for some α > 0 and c > 0: The tail of the distribution
falls off faster than an exponential, i.e., large degrees are unlikely.
− P (d ) = c d −γ , for some γ > 0 and c > 0: Power-law distribution:
The tail of the distribution is fat, i.e., there tend to be many more
nodes with very large degrees.
− Appear in a wide variety of settings including networks describing
incomes, city populations, WWW, and the Internet
− Also known as a scale-free distribution: a distribution that is unchanged
(within a multiplicative factor) under a rescaling of the variable
− Appear linear on a log − log plot
16
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Diameter and Average Path Length

◦ Let l (i, j ) denote the length of the shortest path (or geodesic)
between node i and j (or the distance between i and j).
◦ The diameter of a network is the largest distance between any two
nodes in the network:
diameter = max l (i, j )
i,j

◦ The average path length is the average distance between any two
nodes in the network:
∑i ≥j l (i, j )
average path length = n(n−1)
2
◦ Average path length is bounded from above by the diameter; in some
cases, it can be much shorter than the diameter.
◦ If the network is not connected, one often checks the diameter and
the average path length in the largest component.
17
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Clustering Coefficient

◦ Measures the extent to which my friends are friends with one another.
◦ This clustering measure is represented by the overall clustering
coefficient Cl (G ), given by

3 × number of triangles in the network

Cl (G ) = ,
number of connected triples of nodes
where a “connected triple” refers to a node with edges to an
unordered pair of nodes.
− Note that 0 ≤ Cl (G ) ≤ 1.
− Cl (G ) measures the fraction of triples that have their third edge filled
in to complete the triangle.
− Also referred to as network transitivity: measures the extent that a
friend of my friend is also my friend.

18
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Clustering Coefficient

◦ Another measure of clustering is defined on an individual node basis:

The individual clustering for a node i is
number of triangles connected to vertex i
Cli (g ) = .
number of triples centered at i
1
◦ The average clustering coefficient is Cl Avg (g ) = n ∑i Cli (g ).

Figure: The overall clustering coefficient for this network is 3/8. The individual
clustering for the nodes are 1, 1, 1/6, 0, and 0.

19
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Centrality

◦ A micro measure that captures importance of a node’s position in the

network. Many different centrality measures:

◦ Degree centrality: For node i,

Ci = ki , where ki is the degree of node i.

− For directed networks, both in-degree and out-degree can be used as

centrality measures.
− Simple, but intuitive: individuals with more connections have more
influence and more access to information.
− Does not capture “cascade effects”: importance better captured by
having connections to important nodes
− for example, eigenvector centrality which we shall study soon

20
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Centrality

Figure: Degree centrality of Nodes 2,. . . , d are the same in both graphs.

21
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Centrality

◦ Closeness centrality: Tracks how close a given node is to any other node:
For node i, one such measure is

! −1
1
n − 1 j∑
Ci = dij
6 =i

where dij is the distance between i and j.

◦ Nodes which are close to other nodes on average have high centrality: such
nodes may have more direct influence on others.

22
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Centrality

◦ Closeness centrality: An Example

◦ Network of movie actors: two actors are connected if they work together
◦ Highest centrality 0.4143 for Christopher Lee
− Entered into the Guinness Book of World Records in 2007 for most
screen credits
◦ Lowest centrality 0.1154 for Leia Zanganeh
− An Iranian Theatre and Film Actress.

◦ Limitations
− Spans a very small range.
− For disconnected networks, leads to zero centrality for all nodes!

23
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Centrality

◦ Betweenness centrality: Measures the extent to which a node lies on paths

between other nodes.

i be the number of shortest paths between nodes s to t that pass

◦ Let nst
through a node i
◦ Let gst be the total number of shortest paths from s to t.
◦ Then, betweenness centrality of node i is defined as
i
nst
Ci = ∑ .
s ,t gst

◦ Nodes with high betweenness centrality may have high influence since they
control information passing between others.

24
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Centrality

◦ Betweenness centrality differs from others: it is not a measure of how

well-connected a node is.
◦ Moreover, it spans a wide range.
◦ Let us consider our example of Network of actors
◦ Highest centrality 7.47x108 for Fernando Rey
− A Spanish film, theatre, and television actor
− Worked in both Europe and the United States
− Starred in movies like French Connection (US), Triastana (Spanish),
Cet obscur objet du dsir (French), Ese oscuro objeto del deseo
(Spanish)
◦ Lowest centrality 8.91x105 .

25
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Centrality

◦ And before we forget, in the network of actors

− Highest degree centrality 98734 for Bess Flowers
− She was an American actress best known for her work as an extra in
hundreds of films
◦ So which of these centrality measure is really useful?
− Well, it depends on what is the “objective”
◦ In a friendship network, degree centrality would correspond to who is the
most popular kid. This might be important for certain questions.
◦ Closeness centrality would correspond to who is closest to the rest of the
group, so this would be relevant if we wanted to understand who to inform
or influence for information to spread to the rest of the network.
◦ Betweenness would be relevant if the thought experiment was which
individuals would have to be taken out of the network in order to break the
network into separate clusters.
26
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Centrality and Power in a Network

◦ One application of this is to intermarriage network of Florentine families.

◦ The Medicis emerged as the most influential family in 15th century Florence.
Cosimo de Medici ultimately formed the most politically powerful and
economically prosperous family in Florence, dominating Mediterranean trade.
◦ The Medicis, to start with, were less powerful than many other important
families, both in terms of political dominance of Florentine institutions and
economic wealth.
◦ How did they achieve their prominence?
◦ It could just be luck (in social science, we have to be very careful to
distinguish luck from a systematic pattern, and correlation from causation).
◦ An interesting explanation, eschewing luck, is offered by Padgett and Ansell
(1993) “Robust Action and the Rise of the Medici”— they were the most
powerful family because of their high betweenness centrality, which meant
that they were part of many deals between families supported by marriage
linkages.
27
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Centrality and Power in a Network

Figure: Political and friendship blockmodel structure (Padgett and Ansell 1993)
Image by MIT OpenCourseWare. Adapted from Figure 1.1 on p. 4 in Jackson, Matthew O.Social and Economic Networks. Princeton, NJ: Princeton University Press,
2008.ISBN-13: 9780691134406. ISBN-10: 0691134405.

28
 Networks: Lectures 2 & 3 Graphs Measurements, Metrics

Centrality and Power in a Network

◦ It turns out that Medicis has a very high betweenness centrality 0.522.

◦ No other family has betweenness centrality greater than 0.255.

◦ So the Medicis may have played a central role in holding the network
of influential families in Florence together and thus gained “power”
via this channel.

29
MIT OpenCourseWare
https://ocw.mit.edu

14.15J/6.207J Networks
Spring 2018

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.