Modelling, Mining, and Searching Networks

Networks - Bonato 1

Modelling, Mining, and Searching Networks

Anthony BonatoRyerson University

Master’s SeminarNovember 2012

Networks - Bonato 2

21st Century Graph Theory:Complex Networks

• web graph, social networks, biological networks, internet networks, …

Networks - Bonato 3

• a graph G = (V(G),E(G)) consists of a nonempty set of vertices or nodes V, and a set of edges E

nodes edges• directed graphs (digraphs)

Networks - Bonato 4

Degrees• the degree of a node x, written

deg(x)is the number of edges incident with x

First Theorem of Graph Theory:

V(G)x

|E(G)|2deg(x)

Networks - Bonato 5

The web graph

• nodes: web pages

• edges: links

• over 1 trillion nodes, with billions of nodes added each day

Networks - Bonato 6

Ryerson

GreenlandTourism

Frommer’s

Four SeasonsHotel

City of Toronto

Nuit Blanche

Networks - Bonato 7

Small World Property

• small world networks introduced by social scientists Watts & Strogatz in 1998– low distances

between nodes

Networks - Bonato 8

Power laws in the web graph• power law degree distribution

(Broder et al, 01)

2 some ,, bniN bni

Networks - Bonato 9

Geometric models• we introduced a

stochastic network model which simulates power law degree distributions and other properties– Spatially Preferred

Attachment (SPA) Model

• nodes have a region of influence whose volume is a function of their degree

Networks - Bonato 10

SPA model (Aiello,Bonato,Cooper,Janssen,Prałat, 09)

• as nodes are born, they are more likely to enter a region of influence with larger volume (degree)

• over time, a power law degree distribution results

Networks - Bonato 11

Networks - Bonato 12

Biological networks: proteomics

nodes: proteins

edges: biochemical

interactions

Yeast: 2401 nodes11000 edges

Networks - Bonato 13

Protein networks• proteins are essential

macromolecules of life• understanding their

function and role in disease is of importance

• protein-protein interaction networks (PPI)– nodes: proteins– edges: biochemical

interaction

Networks - Bonato 14

Domination sets in PPI (Milenkovic, Memisevic, Bonato, Przulj, 2011)• dominating sets in graphs

• we found that dominating sets inPPI networks are vital for normalcellular functioning and signalling

– dominating sets capture biologically vital proteins and drug targets– might eventually lead to new drug therapies

Networks - Bonato 15

Social Networks

nodes: people

edges: social interaction(eg friendship)

Networks - Bonato 16

On-line Social Networks (OSNs)Facebook, Twitter, LinkedIn, Google+…

Networks - Bonato 17

Lady Gaga is the centre of Twitterverse

Dalai Lama

Lady Gaga

Anderson Cooper

Queen Rania of Jordan

Arnold Schwarzenegger

Networks - Bonato 18

6 degrees of separation

• Stanley Milgram: famous chain letter experiment in 1967

Networks - Bonato 19

6 Degrees in Facebook?• 1 billion users, > 70

billion friendship links• (Backstrom et al., 2012)

– 4 degrees of separation in Facebook

– when considering another person in the world, a friend of your friend knows a friend of their friend, on average

• similar results for Twitter and other OSNs

Networks - Bonato 20

Dimension of an OSN• dimension of OSN: minimum number of

attributes needed to classify nodes

• like game of “20 Questions”: each question narrows range of possibilities

• what is a credible mathematical formula for the dimension of an OSN?

Networks - Bonato 21

GEO-P model (Bonato, Janssen, Prałat, 2012)

• reverse engineering approach– given network data GEO-P model predicts dimension

of an OSN; i.e. the smallest number of attributes needed to identify users

• that is, given the graph structure, we can (theoretically) recover the social space

Networks - Bonato 22

6 Dimensions of Separation

OSN Dimension

YouTube 6Twitter 4Flickr 4

Cyworld 7

Cops and Robbers

Networks - Bonato 23

C

C

C

R

Cops and Robbers

Networks - Bonato 24

C

C

C

R

Cops and Robbers

Networks - Bonato 25

C

C

C

R

cop number c(G) ≤ 3

Cops and Robbers• played on reflexive undirected graphs G• two players Cops C and robber R play at alternate

time-steps (cops first) with perfect information• players move to vertices along edges; allowed to

moved to neighbors or pass • cops try to capture (i.e. land on) the robber, while

robber tries to evade capture• minimum number of cops needed to capture the

robber is the cop number c(G)– well-defined as c(G) ≤ |V(G)|

Networks - Bonato 26

Networks - Bonato 27

Applications of Cops and Robbers

• moving target search– missile-defense– gaming

• counter-terrorism– intercepting messages or agents

How big can the cop number be?

• if the graph G with order n is disconnected, then the cop number can be as n

• if G is connected, then no one knows how big the cop number can be!

• Meyniel’s Conjecture: c(G) = O(n1/2).

Networks - Bonato 28

Networks - Bonato 29

Networks - Bonato 30

Example of a variantThe robber fights back!

• robber can attack neighbouring cop

• one more cop needed in this graph (check)• Conjecture: For any graph with this modified game, one

more cop needed than for usual cop number.

C

C

C

R

Networks - Bonato 31

Thesis topics• what precisely is a community in a complex

network? • biological network models

– more exploration of dominating sets in PPI• fit GEO-P model to OSN data

– machine learning techniques• new models for complex networks• Cops and Robbers games

– Meyniel’s conjecture, random graphs, variations: good vs bad guy games in graphs

Good guys vs bad guys games in graphs

32

slow medium fast helicopter

slow traps, tandem-win

medium robot vacuum Cops and Robbers edge searching eternal security

fast cleaning distance k Cops and Robbers

Cops and Robbers on disjoint edge sets

The Angel and Devil

helicopter seepage Helicopter Cops and Robbers, Marshals, The Angel and Devil,Firefighter

Hex

badgood

Networks - Bonato

Networks - Bonato 33

Brief biography• over 80 papers, two books, two edited proceedings, with

40 collaborators (many of which are my students)• over 250K in research funding in past 6 years

– grants from NSERC, Mprime, and Ryerson• supervised 8 masters students, 2 doctoral, and 7 post-

docs• over 30 invited addresses world-wide (India, China,

Europe, North America)• won 2011 and 2009 Ryerson Research awards• editor-in-Chief of journal Internet Mathematics; editor of

Contributions to Discrete Mathematics

Networks - Bonato 34

AM8204 – Topics in Discrete Mathematics

• Winter 2012• 6 weeks each: complex networks, graph

searching• project based• Prequisite: AM8002 (or permission from

me)

Networks - Bonato 35

Graphs at Ryerson (G@R)