alexander mikov - program tools for dynamic investigation of social networks

48
Authors: Alexander Mikov, Elena Zamyatina,Daria Germanova 1 Cuban State University, Cuban State University, Higher School Of Economics (Perm Branch) Higher School Of Economics (Perm Branch) , , Perm State Univerity Perm State Univerity Program Tools for Dynamic Investigation of Social Networks

Upload: aist

Post on 16-Jan-2017

93 views

Category:

Data & Analytics


0 download

TRANSCRIPT

Page 1: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Authors:Alexander Mikov, Elena Zamyatina,Daria Germanova

1

Cuban State University, Cuban State University, Higher School Of Economics (Perm Branch)Higher School Of Economics (Perm Branch),,Perm State UniverityPerm State Univerity

Program Tools for Dynamic Investigation of Social Networks

Page 2: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Higher School Of EconomicsNational Research University

Perm Branch

2

Page 3: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Outline

• BackGround• Related Works• Specific Properties of Simulation Software • Simulation Model Representation• The Algorithm of Investigation• Graphs

3

Page 4: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Background

• Social Network Analyses:– Telecommunication.– Marketing.– Sociology.– Etc.

• Information

4

Page 5: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

5

Page 6: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Related Works

• NetLogo• Repast (Recursive Porous Agent

Simulation Toolkit)• Mason

6/52

Page 7: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Winter Simulation Conference • Jonathan K. Alt, Stephen Lieberman. Representing

Dynamic Social Networks In Discrete Event Social Simulation (Modeling, Virtual Environments and Simulation (MOVES) Institute, Naval Postgraduate School, Monterey, California 93943, USA

• Gatti et al. A Simulation-based Approach To Analyze The Information Diffusion In Microblogging Online Social Network. IBM Research-Brazil Av. Tut´oia 1157, San Paulo (SP), BRAZIL

• O¨ zgu¨r O¨ zmen et al. An Agent-based Simulation Study Of A Complex Adaptive Collaboration Network. Oak Ridge National Laboratory. Oak Ridge, TN 37830, USA

7/52

Page 8: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Requirements:

8

Page 9: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Simulation system Triad and it’s purposes in past

• Triad – 80-90 years of 20-th century• Linguistic and software tools for computer

aided design of computer systems (aviation industry)

9

Page 10: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Triad.Net – the Distributed Simulation System with a remote access

10

TriadClientTriadBuilder,TriaDebuggerTriadSecurity

Web-browser

TriadEditor,TriadBuilder,

TriadDebuggerTriadSecurity

Data Base

Server

Server

Simulation Model

TriadBalance

Page 11: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Triad.Net components

11

TriadEditor-the subsystem for Triad-model design

TriadDebugger – Triad verification and validation

TriadMining – the subsystem of an output data intellectual processing

TriadSecurity – the safety subsystem

TriadCompile-Triad-compiler

TriadBuilder-the subsystem of completeness of partly defined model

TriadCore-the Kernel of Triad simulation system

TriadBalance – the subsystem with load balancing function

Page 12: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Simulation model description

• Structure layer (STR)

• Routine layerRoutine layer ( (ROUTROUT))• Message layer (MES)Message layer (MES)

• The layer of structure is dedicated to the description of the objects of the simulation model and their interconnections

12

Page 13: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Simulation model description

• Structure layerStructure layer ((STRSTR))..

• Routine layer (ROUT).

• Message layerMessage layer ((MESMES))..

• The layer of routines describes the behavior of the simulation model objects

13

Page 14: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Simulation model description

• Structure layerStructure layer ((STRSTR))..

• Routine layerRoutine layer ((ROUTROUT))..

• Message layer (MES).

• The layer of messages is used for description

of the messages with complicated structures

14

Page 15: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Structure layer operations

• To add/to delete the poles, nodes and arcs.• The operations with graphs (union, intersection and so on)• Graph constants using.

15

P[2]

P[1]

P[2]

P[1]

V[1]V[2]

G

G = G + V[2](A)

V[2]

A

+ (V[2].A -> V[1].P[1])open V[1] := Str1 [ P[1]=Outp, P[2]=Inp ]

Outp

Inp

Page 16: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Graph Constants

16

Simple path

Path(m)

dPath(m)

Simple cycle

Cycle(m)

dCycle(m)

Complete

Compl(m)

dCompl(m)

Grid

Rectan(m1,m2…)

dRectan(m1,m2…)

Tree

Tree(m,n)

dTree(m,n)

Bipart

Bipart(m,n)

dBipart(m,n)

Star

Star(m)

dStar(m)

Disconnected

nc(m)

dnc(m)

m=4m=6m=4m1=2; m2=3m=3; n=2m=2; n=3m=5m=5

Page 17: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Random graph (model of Erdösh-Renyi) with 30 nodes and probability p=0,25.

17/52

Page 18: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Random graph (model of Erdösh-Renyi) with a number of nodes=30 and a probability p=1.

18/52

Page 19: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Barabashi-Albert graph

19/52

Page 20: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

An example: the model of computer network and

it’s description using Triad

20

Router

Router

Router

Router

Router

Type Router,Host; integer i;M:=dStar(Rout[5]<Pol[4]>);M:=M+node Hst[8]<Pol>;M.Rout[0]=>Router;for i:=1 by 1 to 4 do

M.Rout[i]=>Router;M:=M+edge(Rout[i].Pol[1]—

Hst[2*i-2]);M:=M+edge(Rout[i].Pol[2]—

Hst[2*i-1]);endf;for i:=0 by 1 to 7 do

M.Hst[i]=>Host;endf;

Page 21: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Graphical Editor

21

Page 22: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

The Algorithm of Investigation

22

Min(var)

Count(event)

Check(Pol)

Information procedureSimulation Model

Conditions of simulation

Page 23: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

A List of Standard Information Procedures in Graphical Editor

23

Page 24: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

An example of information procedure –intellectual analyze

information procedure EVENT_SEQUENCE (in ref event E1,E2,E3;out Boolean ARRIVED)

initial interlock (E2,E3); ARRIVED := false; case of E1:available(E2); E2:available( E3); E3:ARRIVED:=true; endc endinf

24An investigator may detect the arrival of the sequence of events E1→E2→E3

Page 25: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Structural Characteritics

25/52

Page 26: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Conditions of simulation

26

Conditions of simulation<name>(<a list of generic parameters>)

(<input and output formal parameters>) initial <a sequence of statements> endi

<a list of information procedures> <a sequence of statements>

processing <a sequence of statements>…endprocendcond

Page 27: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Simulation run

27

simulatesimulate

<<a list of an elements of models, being inspected> on conditions of simulation <name>

(a list of actual generic parameters>)[<a list of input and output actual parameters>]

Page 28: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

28

Page 29: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Distributed simulation model

• Distributed simulation models – a set of logical processes being fulfilled on different compute nodes and communicating with one another by passing messages;

• Each Logical process has local time calendar

• Time paradox – arriving input message with time stamp less than local time

29

Page 30: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

N1

N4

N3

N2

N6N5

N1

N4

N3

N2

N6N5

LP1

LP2

Distributed Models

30LP1 LP2 LPN

Time Time paradox

Page 31: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Two classes of algorithms

• Conservative

31

Optimizations - lookahead

Page 32: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Optimistic Algorithm

32

Problems: •Rollbacks•Memory for states

Page 33: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

33

To find out the hidden dependences of events

Simulation model structure

The sequence of accumulated events

The results of the former runs

The expert knowledge

Page 34: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Expirements

34

ILLIAC

050

100150200250300350400

0 20 40 60 80

количество вычислительных узлов

врем

я вы

полн

ения

Оптимистический АлгоритмКонсервативный алгоритмАлгоритм TriadRule

Page 35: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Experiments

35

Откаты

020406080

100120140160

0 20 40 60 80

количество узлов

коли

чест

во о

ткат

ов

Оптимистический алгоритм

Алгоритм TriadRule

Page 36: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Thank you for your kindly Attention

36

Page 37: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Load balancing

37

Page 38: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

The reasonses of disbalance

38

Communication lines geterogenity

Computing nodes geterogenity

Distributed application geterogenity

Simulation Model Operations

Page 39: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

39

Node1(Head Node)

Node4(Compute Node)

Node2(Compute Node)

Node5(Compute Node)

Node3(Compute Node)

Node6(Compute Node)

Object 1 Object 2

Object 3 Object 4

Object 8 Object 9

Object 6

Object 10

Object 7

Object 5

Node1(Head Node)

Node4(Compute Node)

Node2(Compute Node)

Node5(Compute Node)

Node3(Compute Node)

Node6(Compute Node)

Object 1 Object 2

Object 3 Object 4

Object 8Object 9

Object 6

Object 10

Object 7

Object 5

Page 40: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Decentralized algorithm multiagent load balancing subsystem

40

Compute node

Cluster

Data baseRemote Access

Page 41: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Computing Node and a Simulation Model Fragment

41

Monitoring Agent

(SM)

Monitoring Agent

(CS)

Agent of Analyses

Agent of distribution

Agent of migration

Page 42: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Communication of the agents

42

Agent of migrationAgent of distribution

Black Board Black Board

Monitoring Agent Monitoring Agent

Agent of AnalysesAgent of

distribution

Page 43: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Rules and metarules for the agent of distribution

MetaRule

MetaRule1 MetaRule2 MetaRuleN

1Rule 2Rule 3Rule RuleK 1+RuleK

43

Page 44: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

44

The procedure for the frequency of specific event monitoring

infprocedure EventAverage(event V; in real T ) : realinitial

integer eventCount;real time;eventCount := 0;time := 0;

endihandling

time := T;eventCount:= eventCount + 1;

endhprocessing

EventAverage := eventCount / time;endp

endinf

Page 45: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

45

Subsystem of visualization of the results of monitoring

Page 46: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Conclusion

• Multimodel investigation of computer networks

• Adaptability of software to incorporate into a simulation model new devices and new algorithms that govern their work

• Simulation Model Completeness Analyses and Simulation model Redefining

46

Page 47: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Conclusions

47

Page 48: Alexander Mikov - Program Tools for Dynamic Investigation of Social Networks

Thank you for your kindly Attention

48