+ All Categories
Home > Technology > P2P EC: A study of viability

P2P EC: A study of viability

Date post: 28-Aug-2014
Category:
Upload: juan-luis-jimenez-laredo
View: 652 times
Download: 4 times
Share this document with a friend
Description:
 
Popular Tags:
55
Introduction Background Complex Networks Newscast protocol Model Design The Evolvable Agent Model Properties Experimental Analysis Goals Methodology Analysis of Results Test-Case 1 Test-Case 2 Test-Case 3 Conclusions Peer-to-Peer Evolutionary Computation A Study of Viability Juan Luis Jim´ enez Laredo Dpto. Arquitectura y Tecnolog´ ıa de Computadores Universidad de Granada 27 de Mayo 2010 1 / 44
Transcript
Page 1: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Peer-to-Peer Evolutionary ComputationA Study of Viability

Juan Luis Jimenez Laredo

Dpto. Arquitectura y Tecnologıa de ComputadoresUniversidad de Granada

27 de Mayo 2010

1 / 44

Page 2: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Scope

• Status: Peer-to-Peer Evolutionary Computation (P2P EC)represents a parallel solution for hard problemsoptimization

• Objective: Find empirical evidences showing the viabilityof the P2P EC paradigm

• Modelling: Fine grained parallel EA using a P2P protocolas underlying population structure

2 / 44

Page 3: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Outline

1 Introduction

2 BackgroundComplex NetworksNewscast protocol

3 Model DesignThe Evolvable AgentModel Properties

4 Experimental AnalysisGoalsMethodologyAnalysis of Results

Test-Case 1Test-Case 2Test-Case 3

5 Conclusions

3 / 44

Page 4: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Introduction

EAs: Bio-inspired population based optimization methods

4 / 44

Page 5: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Introduction

5 / 44

Page 6: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Introduction

5 / 44

Page 7: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Introduction

5 / 44

Page 8: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Introduction

6 / 44

Page 9: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Introduction

6 / 44

Page 10: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Introduction

6 / 44

Page 11: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Introduction

P2P EC

• Virtualization:Single view atapplication level

• Decentralization:No centralmanagement

• Massive Scalability:Up to thousands ofcomputers

7 / 44

Page 12: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Outline

1 Introduction

2 BackgroundComplex NetworksNewscast protocol

3 Model DesignThe Evolvable AgentModel Properties

4 Experimental AnalysisGoalsMethodologyAnalysis of Results

Test-Case 1Test-Case 2Test-Case 3

5 Conclusions

8 / 44

Page 13: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Population Structure as a complex network

Watts-Strogatz Model• Easy model for constructing small-world networks

• Begins with a ring

• Rewired edges at random with a probability p

p = 0: Ring p = 0.2: Small-world p = 1: Random

9 / 44

Page 14: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Population Structure as a complex network

Panmictic Small-world Regular lattice

n(n−1)2

log(n) n

10 / 44

Page 15: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Population Structure as a complex network

Panmictic Small-world Regular lattice

n(n−1)2

log(n) n

10 / 44

Page 16: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Population Structure as a complex network

Panmictic Small-world Regular lattice

n(n−1)2

log(n) n

10 / 44

Page 17: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Population Structure as a complex network

Panmictic Small-world Regular lattice

n(n−1)2

log(n) n

10 / 44

Page 18: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Outline

1 Introduction

2 BackgroundComplex NetworksNewscast protocol

3 Model DesignThe Evolvable AgentModel Properties

4 Experimental AnalysisGoalsMethodologyAnalysis of Results

Test-Case 1Test-Case 2Test-Case 3

5 Conclusions

11 / 44

Page 19: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Newscast

Basic Working Principles• Decentralized P2P protocol

• Every node has a cache acting as a routing table

• Dynamical self-organized network

Jelasity,02

12 / 44

Page 20: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Newscast

Basic Working Principles• Decentralized P2P protocol

• Every node has a cache acting as a routing table

• Dynamical self-organized network

Jelasity,02

12 / 44

Page 21: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Newscast: Bootstrapping and Convergence

Experiment• Different network initializations: Watts-Strogatz and Random

• Network characterization: Average path length, Clustering coefficient

Jelasity,02

13 / 44

Page 22: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Newscast: Robustness

Experiment

• System degradation: Up to 100%. Newscast, Random graph

• Network characterization: Size of largest cluster, Number of partitions

Jelasity,02

14 / 44

Page 23: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Newscast: Scalability

Experiment• System traffic: Sizes of networks 1000 and 10000

• Network characterization: Probability of requests to a node

Jelasity,02

15 / 44

Page 24: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Outline

1 Introduction

2 BackgroundComplex NetworksNewscast protocol

3 Model DesignThe Evolvable AgentModel Properties

4 Experimental AnalysisGoalsMethodologyAnalysis of Results

Test-Case 1Test-Case 2Test-Case 3

5 Conclusions

16 / 44

Page 25: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

The Evolvable Agent Model

Design principles• Agent based approach

• Fine grain parallelization

• Spatially structured EA

• Local selection

17 / 44

Page 26: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

The Evolvable Agent Model

Design principles• Agent based approach

• Fine grain parallelization

• Spatially structured EA

• Local selection

17 / 44

Page 27: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Outline

1 Introduction

2 BackgroundComplex NetworksNewscast protocol

3 Model DesignThe Evolvable AgentModel Properties

4 Experimental AnalysisGoalsMethodologyAnalysis of Results

Test-Case 1Test-Case 2Test-Case 3

5 Conclusions

18 / 44

Page 28: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Multi-threading performance on a local computer

Experiment• Scalability of virtual nodes running in a SMP desktop computer

• Fitness evaluation cost Tf ∈ [0.01 . . . 1] seconds

• Test-bed: 1 processor machine and a dual-core processor machine

ThroughputEA = evaluationstime

Speedup =ThroughputEvAg

Throughputsequential

Speedup =Timesequential

TimeEvAg

Linear speedup up to the number of processors19 / 44

Page 29: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Parallel performance on a P2P infrastructure

Experiment

• Scalability for a evaluation cost of Lζ=[1.5,2,3]

• L ∈ [1 . . . 100]

• N = L2 then N ∈ [1 . . . 10000]

Speedup =NTf

Tp

Tp = Tf + Tcomm + Tlat

Gagne, 03

Linear speedups for demanding evaluation functions20 / 44

Page 30: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Outline

1 Introduction

2 BackgroundComplex NetworksNewscast protocol

3 Model DesignThe Evolvable AgentModel Properties

4 Experimental AnalysisGoalsMethodologyAnalysis of Results

Test-Case 1Test-Case 2Test-Case 3

5 Conclusions

21 / 44

Page 31: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Goals and Test-Cases

Goals

1 Scalability: Suitability for tackling large probleminstances.

2 Fault-tolerance: Suitability for tolerating the systemdegradation.

Test-Cases

• Test-Case 1: Scalability against canonical approaches infailure-free environments

• Test-Case 2: Scalability against other populationstructures in failure-free environments

• Test-Case 3: Fault-tolerance of the model under churn

22 / 44

Page 32: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Outline

1 Introduction

2 BackgroundComplex NetworksNewscast protocol

3 Model DesignThe Evolvable AgentModel Properties

4 Experimental AnalysisGoalsMethodologyAnalysis of Results

Test-Case 1Test-Case 2Test-Case 3

5 Conclusions

23 / 44

Page 33: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Generalised l-trap function

• l-trap function (Ackley,1987):

• 2-trap: not-deceptive• 3-trap: partially

deceptive• 4-trap: fully deceptive

• L = 12 . . . 60

24 / 44

Page 34: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Experimental settings

• Population size• Estimated by bisection• Selectorecombinative

GA (Mutation less)• Minimum population

size able to reach 0.98of SR

• Uniform Crossover

• Binary Tournament

25 / 44

Page 35: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Outline

1 Introduction

2 BackgroundComplex NetworksNewscast protocol

3 Model DesignThe Evolvable AgentModel Properties

4 Experimental AnalysisGoalsMethodologyAnalysis of Results

Test-Case 1Test-Case 2Test-Case 3

5 Conclusions

26 / 44

Page 36: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 1: Scalability

• Failure-free environment

• Canonical approaches: SSGA, GGA

• Metrics: Population Size, Evaluations

27 / 44

Page 37: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 1: Scalability

Settings

Problem instance: 2-trapPop. Size: Tuning AlgorithmNo Mutation

28 / 44

Page 38: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 1: Scalability

Settings

Problem instance: 3-trapPop. Size: Tuning AlgorithmNo Mutation

29 / 44

Page 39: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 1: Scalability

Settings

Problem instance: 4-trapPop. Size: Tuning AlgorithmNo Mutation

30 / 44

Page 40: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 1: Scalability

Larger instance 4-Trap: L=36

Pop. Size: 600Max. Eval: 393000

30 / 44

Page 41: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 1: Scalability

Settings

Equally parameterized SSGA, GGA and EvAgProblem instance: L=36 4-trapPop. Size: 600Max. Eval: 393000Mutation: Bit-flip Pm = 1

L

31 / 44

Page 42: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 2: Population Structure

Ring Watts-Strogatz Newscast

32 / 44

Page 43: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 2: Population Structure

Settings

Problem instance: 2-trapPop. Size: Tuning AlgorithmNo Mutation

33 / 44

Page 44: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 2: Population Structure

Settings

Problem instance: 3-trapPop. Size: Tuning AlgorithmNo Mutation

34 / 44

Page 45: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 2: Population Structure

Settings

Problem instance: 4-trapPop. Size: Tuning AlgorithmNo Mutation

35 / 44

Page 46: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 2: Population Structure

Settings

Equally parameterized approaches using different pop.structuresProblem instance: L=36 4-trapPop. Size: 600Max. Eval: 393000Mutation: Bit-flip Pm = 1

L

36 / 44

Page 47: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 3: Fault-tolerance

37 / 44

Page 48: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 3: Fault-tolerance

37 / 44

Page 49: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 3: Fault-tolerance

• Stutzbach and Rajaie,2006

• Weibull distribution:• X = λ(−ln(U))

1k

• Shape: k = 0.4

• Scale: λ = 400, 2500

38 / 44

Page 50: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 3: Fault-tolerance

Settings

Problem instance: 2-trapPop. Size: Tuning AlgorithmNo Mutation

39 / 44

Page 51: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 3: Fault-tolerance

Settings

Problem instance: 3-trapPop. Size: Tuning AlgorithmNo Mutation

40 / 44

Page 52: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 3: Fault-tolerance

Settings

Problem instance: 4-trapPop. Size: Tuning AlgorithmNo Mutation

41 / 44

Page 53: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Test-Case 3: Fault-tolerance

Settings

Equally parameterized approaches with and without churnProblem instance: L=36 4-trapPop. Size: 600Max. Eval: 393000Mutation: Bit-flip Pm = 1

L

42 / 44

Page 54: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Conclusions

Selected publicationsPeer reviewed journal papers :

1 J.L.J. Laredo, A.E. Eiben, M. van Steen, and J.J. Merelo. Evag: A scalablepeer-to-peer evolutionary algorithm. GPEM, 2010.http://dx.doi.org/10.1007/s10710-009-9096-z.

2 J.L.J. Laredo, P.A. Castillo, A.M. Mora, J.J. Merelo, and C. Fernandes.Resilience to churn of a peer-to-peer evolutionary algorithm. IJHPSA,1(4):260-268, 2009.

3 J.L.J. Laredo, P.A. Castillo, A.M. Mora, and J.J. Merelo. Evolvable agents, afine grained approach for distributed evolutionary computing. SoftComputing, 12(12):1145-1156, 2008.

Peer reviewed conference papers and book chapters :

1 J.L.J. Laredo, P.A. Castillo, A.M. Mora, J.J. Merelo, A.C. Rosa, and C.Fernandes. Evolvable agents in static and dynamic optimization problems. InPPSN X, pages 488-497. Springer, 2008

2 J.L.J. Laredo, A.E. Eiben, M. van Steen, P.A. Castillo, A.M. Mora, and J.J.Merelo. P2P evolutionary algorithms: A suitable approach for tackling largeinstances in hard optimization problems. In Euro-Par’ 08, pages 622-631.Springer, 2008.

3 J.L.J. Laredo, P.A. Castillo, A.M. Mora, and J.J. Merelo. Exploringpopulation structures for locally concurrent and massively parallelevolutionary algorithms. In IEEE WCCI2008 Proceedings, pages 2610-2617.IEEE Press, Hong Kong, June 2008.

43 / 44

Page 55: P2P EC: A study of viability

Introduction

Background

ComplexNetworks

Newscastprotocol

Model Design

The EvolvableAgent

ModelProperties

ExperimentalAnalysis

Goals

Methodology

Analysis ofResults

Test-Case 1

Test-Case 2

Test-Case 3

Conclusions

Questions

Thanks for your attention!

44 / 44


Recommended