Mathematics of Networks

Review on Graphs

MA 653: Network Science

Instructor: Ashok Singh Sairam
[email protected]
Representing networks
• The network consists of entities connected with each other
• The structure of these connections is represented through graphs
• A graph is represented by two sets
• A vertex set V of the entities participating in the network. Let n be the
number of vertices
• Also called node or actor set
• An edge set E of the connections between vertices. Let m will be the number
of edges
• Also called link or tie set

MA 653: Network Science 2

Graph terminology
• A graph whose vertices are connected by at most one link is called
simple network or simple graph
• Most of the graphs we will examine will be simple
• When two nodes connect with more than one edge, we refer to all
those edges collectively as multiedge
• The corresponding graph is called multigraph
• Depending on the type of connection a node might be connected to
itself
• Self-edges or self-loops

MA 653: Network Science 3

Representing graphs
• If we label the nodes with IDs 1, 2, … n we can denote each edge as a
pair (i,j)
• This is an edge list specification
• Good for storing and processing networks in computers, but not for
mathematical development
• The adjacency matrix A of a simple graph is a matrix with elements Aij
such that:

MA 653: Network Science 4

Representing Konigsberg bridges
• Draw the corresponding graph
• Represent the graph using adjacency matrix and adjacency list

MA 653: Network Science 5

Observations
• The diagonal of the adjacency matrix is all 0 => ??

• How does self-loop affect the adjacency matrix?

• The adjacency matrix is symmetric => ??

MA 653: Network Science 6

Weighted networks
• Some relations are not simple on/off (1/0) relations

• In a weighted network links can have weights

• The corresponding adjacency matrix entry is equal to the weight
• Weights can represent the frequency of contacts between the actors, the
capacity of a channel connecting two routers etc.

• When weights are integer it might be convenient to think of the

weight as multiedges

MA 653: Network Science 7

Directed networks
• In some phenomena the direction of the underlying relation between
two nodes matters  relations are not reciprocal
• E.g., Twitter connections, world wide web links, paper citations etc.
• These relations are captured through directed networks/graphs
• The adjacency matrix of a directed graph (or a digraph) is given by:
A is in general not symmetric

Self edges are

represented by setting Aii=1

MA 653: Network Science 8

Ex: Citation network
• Paper 5 cites Paper 4 and Paper 3.
• Paper 4 cites Paper 2.
• Paper 3 cites Paper 2.
• Paper 2 cites no other papers.
• Paper 1 cites no other papers.

MA 653: Network Science 9

Directed Acyclic Graph/Networks
• A cycle in a directed network is a closed loop of edges with the
arrows on each of the edges pointing the same way around the loop
• Networks without cycles are called acyclic, while those with cycles are
called cyclic
• Self-edges also count as cycles
• An acyclic network can be drawn with all edges pointing downward
(not necessarily a unique drawing)
• Example: citation network
• The vertical axis represents date of publication, point from
later to earlier paper

MA 653: Network Science 10

Directed Acyclic Graph/Networks
• DAG construction
• There must be at least one node (node I) with no outgoing edge (Proof ?)
• Remove node I along with its edges (rename it as 1) A B
• Repeat process – find another node with no
outgoing edge, say node H, rename as 2 C
• Finally, redraw the graph D
• edges from higher numbered to lower numbered E
only F

G
I
H

MA 653: Network Science 11

Directed Acyclic Network: construction
• Draw adjacency matrix of the graph?
• What property does the matrix have?

MA 653: Network Science 12

Ex: DAG
• Draw a directed acyclic network with all edges pointing downwards.
• Write the adjacency matrix of the graph.

d
a
b
g
c e

MA 653: Network Science 13

Acyclic network: adjacency matrix
• If the nodes are numbered as
described earlier, the adjaceny matrix
has an interesting property 7 6 5 4 3 2 1
7 0 1 0 1 0 0 0
• all non-zero elements above the diagonal
6 0 0 1 0 0 0 0
• it is upper triangular
5 0 0 0 0 1 0 1
• Diagonal elements are zero 4 0 0 0 0 1 1 0
• Triangular matrix with zeros on the diagonal 3 0 0 0 0 0 0 1
are called strictly triangular 2 0 0 0 0 0 0 1
• All of the eigenvalues of an adjacency 1 0 0 0 0 0 0 0

matrix are zero if and only if the

network is acyclic
MA 653: Network Science 14
Hypergraphs
• Hypergraphs: Simple graph where an edge can
connect more than two vertices
1 4
• Vertex: Similar to conventional graph (person,
gene, protein, etc.)
3
• Hyperedge: An edge is called a hyperedge. A
hyperedge can connect one or more vertices
2 5
• Degree of a vertex: number of hyperedges to which
it belongs
• Application: Social networks, biology network

MA 653: Network Science 15

Representing hypergraph with bipartite graph
• Introduce a new node corresponding to each
group
1 4
• There are 5 groups, so introduce 5 new
nodes
3

2 5

1 2 3 4 5

MA 653: Network Science 16

Social Network Data: 1-mode
• Social network data are substantively divided by the number of
modes in the data
• 1-mode data represents edges based on direct contact between
actors in the network.
• All the nodes are of the same type (people, organization, ideas, etc).
• Examples: Communication, friendship, giving orders, sending email.

MA 653: Network Science 17

Social Network Data: 2-mode
• 2-mode data represents nodes from two separate classes (actor &
societies), where all ties are across classes.
• Examples:
• Actors as members of societies
• Actors as authors on papers
• Words used often by actors
• Events in the life history of actors
• The two modes of the data represent a duality:
• One way to analyze two-mode data is to create projections onto each mode

MA 653: Network Science 18

Bipartite networks
• In bipartite networks (two-mode networks) there are two kinds of
vertices
• One type represents the original vertices (e.g., people) and the other
represents the groups – or affiliations or societies – that the first type of
vertices belong to (e.g., company)
• Edges of a bipartite network run only between nodes of different type
• A bipartite network is defined by the incidence matrix B. If there are n
number of actors and g number of groups:

MA 653: Network Science 19

Ex: Working with two mode data (Persons)
• What can you say about persons A & F? D & F?
Groups
• Get the overlap for any pair of groups / persons 1 2 3 4 5
A 0 0 0 0 1
• summing the multiplied elements of the B 1 0 0 0 0

Persons
corresponding rows/columns C 1 1 0 0 0
D 0 1 1 1 1
E 0 0 1 0 0
F 0 0 1 1 0

g
Pij   Aik A jk
k 1
MA 653: Network Science 20
Ex: Working with two mode data (Groups)
• Similarly for groups
Similarly for Groups:
Groups 1 2 12
•

1 2 3 4 5 A 0 0 0 Group 1 has 2
A 0 0 0 0 1 B 1 0 0 members,
B 1 0 0 0 0 C 1 1 1 group 2 has 2
Persons

D 0 1 0 members and
C 1 1 0 0 0
they overlap
D 0 1 1 1 1 E 0 0 0
by 1 members
E 0 0 1 0 0 F 0 0 0 (C).
F 0 0 1 1 0 S 2 2 1
p
Gij   Aki Akj
k 1

MA 653: Network Science 21

One-mode projections
• We can use the bipartite network to infer connections between nodes of the same
type
• Projection on the actors n-vertex network,
• Nodes are actors
• Edge between two actors exists if they participate to at least one common
group
• Projection on the groups g-vertex network,
• nodes are the groups
• Edge between two groups exists if they include at least one common actor
• One mode projections discard a lot of the information present in the structure of
the original bipartite graph

MA 653: Network Science 22

Ex: One mode projection
Groups

Groups 1 2 3 4 5
1 2 3 4 5
A 0 0 0 0 1
B 1 0 0 0 0
Persons

C 1 1 0 0 0
D 0 1 1 1 1
A B C D E F
E 0 0 1 0 0
F 0 0 1 1 0

Persons
MA 653: Network Science 23
Mathematics of one-mode projection (1)
• Two actors i and j belong to group k iff BkiBkj=1
• The total number of groups Pij that i and j belong is given by:

• The n x n matrix P=BTB is similar to the adjacency matrix for the

weighted one-mode projection onto the n vertices
• Why only similar and not the same?

MA 653: Network Science 24

 1 2 3 4 5 A B C D E F
A 0 0 0 0 1 1 0 1 1 0 0 0
B 1 0 0 0 0 2 0 0 1 1 0 0
A= C 1 1 0 0 0 A = 3
T 0 0 0 1 1 1
D 0 1 1 1 1 4 0 0 0 1 0 1
E 0 0 1 0 0 5 1 0 0 1 0 0
F 0 0 1 1 0

(6x5) (5x6)

P = A(AT)
G = AT(A)

MA 653: Network Science 25

 1 2 3 4 5 A B C D E F
A 0 0 0 0 1 1 0 1 1 0 0 0
B 1 0 0 0 0 2 0 0 1 1 0 0
A= C 1 1 0 0 0 AT = 3 0 0 0 1 1 1
D 0 1 1 1 1 4 0 0 0 1 0 1
E 0 0 1 0 0 5 1 0 0 1 0 0
F 0 0 1 1 0

(6x5) (5x6)

P = A(AT)
G = AT(A)
A * AT = P AT * A = P
(6x5)(5x6) (6x6) (5x6) 6x5) (5x5)
P
A B C D E F G
A 1 0 0 1 0 0 1 2 3 4 5
B 0 1 1 0 0 0 1 2 1 0 0 0
C 0 1 2 1 0 0 2 1 2 1 1 1
D 1 0 1 4 1 2 3 0 1 3 2 1
E 0 0 0 1 1 1 4 0 1 2 2 1
F 0 0 0 2 1 2 5 0 1 1 1 2

MA 653: Network Science 26

Mathematics of one-mode projection
• One can get either projection (person-to-person or group-to-group)
easily with a little matrix multiplication
• First define AT as the transpose of A (simply reverse the rows and
columns).
• Person-to-person projection will be given by P= AAT
• Similarly, group-to-group project will be given by G= ATA
• The n x n matrix P=ATA is similar to the adjacency matrix for the
weighted one-mode projection onto the n vertices
• Why only similar and not the same?

MA 653: Network Science 27

Co-citation coupling
• The co-citation of two vertices i and j in a directed network is the
number of nodes that have outgoing edges pointing to both nodes i
and j
• So in order for node k to contribute to the co-citation of i and j, the following
needs to be true: AikAjk=1
• Hence the co-citation of nodes i and j is:

• The n x n adjacency matrix of the corresponding co-citation network

is:
MA 653: Network Science 28
Example: Co-citation coupling

MA 653: Network Science 29

A= AT =

C = AAT 1 2 3 4 5 6
1 2 1 1 0 1 0
2 1 2 0 1 1 0
3 1 0 2 0 2 0
4 0 1 0 2 1 0
5 1 1 2 1 3 0
6 0 0 0 0 0 0

MA 653: Network Science 30

A= AT =

C = AAT 1 2 3 4 5 6
1 2 1 1 0 1 0
2 1 2 0 1 1 0
3 1 0 2 0 2 0
4 0 1 0 2 1 0
5 1 1 2 1 3 0
6 0 0 0 0 0 0

MA 653: Network Science 31

Problem 1
• Consider certain papers and the references that they cite
• 1 {2,4,5}
• 2  {3,4,5}
• 3 {4}
• 4 {}
• Which of these papers do you expect to be likely of similar topic?
Problem 2
• Consider a product website that maintains a purchase profile of the
users over a time period T. A snap shot of the same for 4 users are as
follows:
• A  {1,2,3}
• B  {1,3}
• C  {1,2, 3, 4}
• D { 3,4}

• If A purchases an unknown product (say 5), can you say which other
user will most likely purchase the same product?

33