CS 6604: Data Mining Large Networks and Time-series. B. Aditya Prakash Lecture #8: Epidemics: Thresholds

Similar documents
Toward Understanding Spatial Dependence on Epidemic Thresholds in Networks

CS 6604: Data Mining Large Networks and Time- series. B. Aditya Prakash Lecture #9: Epidemics: Compe3ng Viruses

Virus Propagation on Time-Varying Networks: Theory and Immunization Algorithms

Social Influence in Online Social Networks. Epidemiological Models. Epidemic Process

Epidemics and information spreading

Epidemic Thresholds in Real Networks

Linear Algebra Methods for Data Mining

CS224W: Analysis of Networks Jure Leskovec, Stanford University

ECS 289 / MAE 298, Lecture 7 April 22, Percolation and Epidemiology on Networks, Part 2 Searching on networks

Epidemics in Complex Networks and Phase Transitions

Social and Economic Networks Matthew O. Jackson

Winner Takes All: Competing Viruses or Ideas on fair-play Networks

Epidemic Spread in Mobile Ad Hoc Networks: Determining the Tipping Point

Competing Memes Propagation on Networks: A Case Study of Composite Networks

An Optimal Control Problem Over Infected Networks

Winner Takes All: Competing Viruses or Ideas on fair-play Networks

SIS epidemics on Networks

Spreading and Opinion Dynamics in Social Networks

Introduction to SEIR Models

The Spreading of Epidemics in Complex Networks

ECS 289 F / MAE 298, Lecture 15 May 20, Diffusion, Cascades and Influence

Die-out Probability in SIS Epidemic Processes on Networks

Epidemics on networks

A new centrality measure for probabilistic diffusion in network

Modelling of the Hand-Foot-Mouth-Disease with the Carrier Population

Analytically tractable processes on networks

An Improved Computer Multi-Virus Propagation Model with User Awareness

Announcements. Topics: Work On: - sec0ons 1.2 and 1.3 * Read these sec0ons and study solved examples in your textbook!

Full Diffusion History Reconstruction in Networks

SI j RS E-Epidemic Model With Multiple Groups of Infection In Computer Network. 1 Introduction. Bimal Kumar Mishra 1, Aditya Kumar Singh 2

Improved Bounds on the Epidemic Threshold of Exact SIS Models on Complex Networks

Mathematical Analysis of Epidemiological Models: Introduction

Mathematical Epidemiology Lecture 1. Matylda Jabłońska-Sabuka

arxiv: v2 [cond-mat.stat-mech] 9 Dec 2010

Diffusion of Innovation

Virgili, Tarragona (Spain) Roma (Italy) Zaragoza, Zaragoza (Spain)

Epidemics on networks

Diffusion of information and social contagion

Lecture VI Introduction to complex networks. Santo Fortunato

Preventive behavioural responses and information dissemination in network epidemic models

6.207/14.15: Networks Lecture 7: Search on Networks: Navigation and Web Search

SIR model. (Susceptible-Infected-Resistant/Removed) Outlook. Introduction into SIR model. Janusz Szwabiński

Link Operations for Slowing the Spread of Disease in Complex Networks. Abstract

Epidemic Threshold of an SIS Model in Dynamic Switching Networks

Modeling, Analysis, and Control of Information Propagation in Multi-layer and Multiplex Networks. Osman Yağan

Global Stability of a Computer Virus Model with Cure and Vertical Transmission

Quantitative model to measure the spread of Security attacks in Computer Networks

Research Article Towards the Epidemiological Modeling of Computer Viruses

Oscillatory epidemic prevalence in g free networks. Author(s)Hayashi Yukio; Minoura Masato; Matsu. Citation Physical Review E, 69(1):

On Susceptible-Infected-Susceptible Epidemic Spreading: an Overview of Recent Study

Machine Learning and Modeling for Social Networks

Non-Linear Models Cont d: Infectious Diseases. Non-Linear Models Cont d: Infectious Diseases

GLOBAL DYNAMICS OF A MATHEMATICAL MODEL OF TUBERCULOSIS

GEMF: GENERALIZED EPIDEMIC MODELING FRAMEWORK SOFTWARE IN PYTHON

Network Science & Telecommunications

Kalavakkam, Chennai, , Tamilnadu, INDIA 2,3 School of Advanced Sciences. VIT University Vellore, , Tamilnadu, INDIA

Lecture 1: Graphs, Adjacency Matrices, Graph Laplacian

Spectral Graph Theory for. Dynamic Processes on Networks

WE FOCUS ON A simple continuous-time model for the

Identifying Core Network Structure for Epidemic Simulations

Stability of SEIR Model of Infectious Diseases with Human Immunity

arxiv:cond-mat/ v2 [cond-mat.stat-mech] 24 Apr 2004

arxiv:cond-mat/ v3 [cond-mat.dis-nn] 10 Dec 2003

Diffusion of Innovation and Influence Maximization

Complex networks: an introduction Alain Barrat

Adventures in random graphs: Models, structures and algorithms

Models of Infectious Disease Formal Demography Stanford Summer Short Course James Holland Jones, Instructor. August 15, 2005

Diffusion processes on complex networks

The minimal spectral radius of graphs with a given diameter

Networks as a tool for Complex systems

Epidemic reemergence in adaptive complex networks

Inferring the origin of an epidemic with a dynamic message-passing algorithm

Section 8.1 Def. and Examp. Systems

arxiv: v1 [physics.soc-ph] 7 Jul 2015

A note on modeling retweet cascades on Twitter

Mathematical Modeling and Analysis of Infectious Disease Dynamics

Optimal contact process on complex networks

Generalized Epidemic Mean-Field Model for Spreading Processes Over Multilayer Complex Networks

Modeling Epidemic Risk Perception in Networks with Community Structure

A Note on the Spread of Infectious Diseases. in a Large Susceptible Population

A BINOMIAL MOMENT APPROXIMATION SCHEME FOR EPIDEMIC SPREADING IN NETWORKS

Threshold Conditions in SIR STD Models

Research Article Two Quarantine Models on the Attack of Malicious Objects in Computer Network

MATHEMATICAL MODELS Vol. III - Mathematical Models in Epidemiology - M. G. Roberts, J. A. P. Heesterbeek

Diffusion of Behavior and Equilibrium Properties in Network Games

The death of an epidemic

Quarantine generated phase transition in epidemic spreading. Abstract

Spread of Malicious Objects in Computer Network: A Fuzzy Approach

arxiv:cond-mat/ v2 [cond-mat.stat-mech] 7 Jan 2000

Models of Communication Dynamics for Simulation of Information Diffusion

Dynamics of a network-based SIS epidemic model with nonmonotone incidence rate

A Time Since Recovery Model with Varying Rates of Loss of Immunity

Least Squares Parameter Es.ma.on

Identifying critical nodes in multi-layered networks under multi-vector malware attack Rafael Vida 1,2,3, Javier Galeano 3, and Sara Cuenda 4

Epidemic spreading on heterogeneous networks with identical infectivity

MODELING THE SPREAD OF DENGUE FEVER BY USING SIR MODEL. Hor Ming An, PM. Dr. Yudariah Mohammad Yusof

Comparative analysis of transport communication networks and q-type statistics

Epidemic dynamics and endemic states in complex networks

The Dynamic Properties of a Deterministic SIR Epidemic Model in Discrete-Time

Maximizing the Spread of Influence through a Social Network. David Kempe, Jon Kleinberg, Éva Tardos SIGKDD 03

Transcription:

CS 6604: Data Mining Large Networks and Time-series B. Aditya Prakash Lecture #8: Epidemics: Thresholds

A fundamental ques@on Strong Virus Epidemic? 2

example (sta@c graph) Weak Virus Epidemic? 3

Problem Statement # Infected above (epidemic) below (ex4nc4on) @me Find, a condi4on under which virus will die out exponen4ally quickly regardless of ini4al infec4on condi4on Separate the regimes? 4

Threshold (sta@c version) Problem Statement Given: Graph G, and Virus specs (apack prob. etc.) Find: A condiuon for virus exuncuon/invasion 5

Threshold: Why important? AcceleraUng simulauons ForecasUng ( What-if scenarios) Design of contagion and/or topology A great handle to manipulate the spreading ImmunizaUon Maximize collaborauon.. 6

Outline Q: What is the epidemic threshold? Background Result and IntuiUon (StaUc Graphs) Proof Ideas (StaUc Graphs) Bonus: Dynamic Graphs 7

SIR model: life immunity (mumps) Each node in the graph is in one of three states SuscepUble (i.e. healthy) Infected Removed (i.e. can t get infected again) Prob. β Prob. δ t = 1 t = 2 t = 3 8

Related Work q q q q q q q R. M. Anderson and R. M. May. InfecUous Diseases of Humans. Oxford University Press, 1991. A. Barrat, M. Barthélemy, and A. Vespignani. Dynamical Processes on Complex Networks. Cambridge University Press, 2010. F. M. Bass. A new product growth for model consumer durables. Management Science, 15(5):215 227, 1969. D. ChakrabarU, Y. Wang, C. Wang, J. Leskovec, and C. Faloutsos. Epidemic thresholds in real networks. ACM TISSEC, 10(4), 2008. D. Easley and J. Kleinberg. Networks, Crowds, and Markets: Reasoning About a Highly Connected World. Cambridge University Press, 2010. A. Ganesh, L. Massoulie, and D. Towsley. The effect of network topology in spread of epidemics. IEEE INFOCOM, 2005. Y. Hayashi, M. Minoura, and J. Matsukubo. Recoverable prevalence in growing scale-free networks and the effecuve immunizauon. arxiv:cond-at/0305549 v2, Aug. 6 2003. q H. W. Hethcote. The mathemaucs of infecuous diseases. SIAM Review, 42, 2000. q H. W. Hethcote and J. A. Yorke. Gonorrhea transmission dynamics and control. Springer Lecture Notes in BiomathemaUcs, 46, 1984. q q q q q q J. O. Kephart and S. R. White. Directed-graph epidemiological models of computer viruses. IEEE Computer Society Symposium on Research in Security and Privacy, 1991. J. O. Kephart and S. R. White. Measuring and modeling computer virus prevalence. IEEE Computer Society Symposium on Research in Security and Privacy, 1993. R. Pastor-Santorras and A. Vespignani. Epidemic spreading in scale-free networks. Physical Review LePers 86, 14, 2001. All are about either: Structured topologies (cliques, block-diagonals, hierarchies, random) Specific virus propaga@on models Sta@c graphs 9

Outline Q: What is the epidemic threshold? Background Result and Intui@on (Sta@c Graphs) Proof Ideas (StaUc Graphs) Bonus: Dynamic Graphs 10

How should the answer look like? Answer should depend on: Graph Virus PropagaUon Model (VPM) But how?? β + δ d avg Graph average degree? max. degree? diameter? VPM which parameters? How to combine linear? quadrauc? exponenual? diameter? ( avg max 2 2 β d avg δd ) / d?.. 11

Sta@c Graphs: Main Result [Prakash+, 2011] Informally, For, Ø any arbitrary topology (adjacency matrix A) Ø any virus propagation model (VPM) in standard literature the epidemic threshold depends only 1. on the λ, first eigenvalue of A, and 2. some constant C VPM, determined by the virus propagation model λ C VPM No epidemic if λ * C VPM < 1 12

Our thresholds for some models s = effec4ve strength s < 1 : below threshold Models Effec@ve Strength (s) Threshold (@pping point) SIS, SIR, SIRS, SEIR s = λ. SIV, SEIV s = λ. SI 1I2V1 V2 (H.I.V.) s = λ. β δ βγ δ β1v v2 ( γ +θ ) 2 + β ε 2 ( ε + v ) 1 s = 1 13

Our result: Intui@on for λ Official defini@on: Let A be the adjacency matrix. Then λ is the root with the largest magnitude of the characteristic polynomial of A [det(a xi)]. Un-official Intui@on J λ ~ # paths in the graph k A λ k. u u Doesn t give much intuiuon! A k(i, j) = # of paths i à j of length k 14

Largest Eigenvalue (λ) beper connecuvity higher λ λ 2 λ = N λ = N-1 λ 2 λ= 31.67 λ= 999 N = 1000 N nodes 15

Examples: Simula@ons SIR (mumps) Frac@on of Infec@ons Time @cks Footprint Effec@ve Strength (a) Infection profile (b) Take-off plot PORTLAND graph 31 million CS 6604:DM links, Large Networks 6 million & Time-Series nodes 16

Examples: Simula@ons SIRS (pertusis) Frac@on of Infec@ons Time @cks Footprint Effec@ve Strength (a) Infection profile (b) Take-off plot PORTLAND graph 31 million CS 6604:DM links, Large Networks 6 million & Time-Series nodes 17

Outline Q: What is the epidemic threshold? Background Result and IntuiUon (StaUc Graphs) Proof Ideas (Sta@c Graphs) Bonus: Dynamic Graphs 18

Proof Sketch General VPM structure Model-based λ * C VPM < 1 Topology and stability Graph-based 19

Some trivia Ø first person in the US idenufied as a healthy carrier of the pathogen associated with typhoid fever. Ø infected some 53 people, over the course of her career as a cook! Ø forcibly quaranuned by public health authoriues 20

Two Infected States? SICR: with a carrier AsymtomaUc SymptomaUc Sneezing I1 I2 21

Ingredient 1: Our generalized model Endogenous Transi@ons Suscep@ble Infected Exogenous Transi@ons Endogenous Transi@ons Vigilant 22

Models and more models Model SIR SIS SIRS SEIR.. SICR MSIR SIV SI 1I2V1 V2. Used for Mumps Flu Pertussis Chicken-pox Tuberculosis Measles Sensor Stability H.I.V. 23

Our generalized model Endogenous Transi@ons Suscep@ble Infected Vigilant 24

Special case: SIR Suscep@ble Infected Vigilant 25

Special case: H.I.V. SI I V V 1 2 1 2 Non-terminal Terminal MulUple InfecUous, Vigilant states 26

Ingredient 2: NLDS + Stability View as a NLDS discrete Ume non-linear dynamical system (NLDS) size mn x 1...... Probability vector Specifies the state of the system at Ume t size N (number of nodes in the graph) S I V 27

Ingredient 2: NLDS + Stability View as a NLDS discrete Ume non-linear dynamical system (NLDS) size mn x 1... Non-linear func@on Explicitly gives the evoluuon of system.... 28

Ingredient 2: NLDS + Stability View as a NLDS discrete Ume non-linear dynamical system (NLDS) Threshold à Stability of NLDS 29

Special case: SIR S size 3N x 1 I R S I R = probability that node i is not apacked by any of its infecuous neighbors NLDS 30

Fixed Point 1 1. 0 0. 0 0. State when no node is infected Q: Is it stable? 31

Stability for SIR Stable Unstable under threshold above threshold 32

See paper for full proof General VPM structure Model-based λ * C VPM < 1 Topology and stability Graph-based 33

Outline Q: What is the epidemic threshold? Background Result and IntuiUon (StaUc Graphs) Proof Ideas (StaUc Graphs) Bonus: Dynamic Graphs 34

Dynamic Graphs: Epidemic? DAY (e.g., work) Alternating behaviors adjacency matrix 8 8 35

Dynamic Graphs: Epidemic? NIGHT (e.g., home) Alternating behaviors adjacency matrix 8 8 36

Model Descrip@on Healthy SIS model recovery rate δ infecuon rate β N1 Infected Prob. β X N2 N3 Prob. δ Set of T arbitrary graphs day N night N, weekend.. N N 37

Obvious result No epidemic if λ max β δ <1 #inf. This looks OK BUT Too pessimisuc! Ume 38

Main result: Dynamic Graphs Threshold [Prakash+, 2010] Informally, NO epidemic if eig (S) = < 1 Single number! Largest eigenvalue of The system matrix S 39

NO epidemic if eig (S) = < 1 S = cure rate Prob. β N1 Infected Prob. δ X Healthy N2 infec@on rate N3 N day adjacency matrix night.. N 40

log(frac4on infected) Synthe@c ABOVE Infec@on-profile MIT Reality Mining ABOVE AT AT BELOW BELOW Time 41

Footprint (# infected @ steady state ) Take-off plots Synthe@c MIT Reality Our threshold EPIDEMIC Our threshold EPIDEMIC NO EPIDEMIC NO EPIDEMIC (log scale) 42