Ec awards lecture spring 2008 advances in parameterless evolutionary algorithms
Download
1 / 38

EC Awards Lecture ~ Spring 2008 Advances in Parameterless Evolutionary Algorithms - PowerPoint PPT Presentation


  • 121 Views
  • Uploaded on

EC Awards Lecture ~ Spring 2008 Advances in Parameterless Evolutionary Algorithms. Lisa Guntly André Nwamba Research Advisor: Dr. Daniel Tauritz Natural Computation Laboratory. Evolutionary Algorithms (EAs). User Parameters. Problem. Evolutionary Algorithm (EA). Solution.

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about ' EC Awards Lecture ~ Spring 2008 Advances in Parameterless Evolutionary Algorithms' - traci


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
Ec awards lecture spring 2008 advances in parameterless evolutionary algorithms

EC Awards Lecture ~ Spring 2008Advances in Parameterless Evolutionary Algorithms

Lisa Guntly

André Nwamba

Research Advisor: Dr. Daniel Tauritz

Natural Computation Laboratory


Evolutionary algorithms eas
Evolutionary Algorithms (EAs)

User Parameters

Problem

Evolutionary Algorithm (EA)

Solution


Evolutionary algorithms
Evolutionary Algorithms

Create Initial Population

Evaluate Fitness

No

Select Parents

Termination

Create Offspring

Yes

Solution

Select Survivors

Evaluate Fitness


Motivation
Motivation

  • Parameter specification complicates EAs

    • Expert knowledge required

    • Time-consuming

    • Sub-optimal - optimal parameter values can change during a run



Parameterless eas our approach
Parameterless EAs: Our Approach

  • Completely Parameterless EAs

  • Biological metaphors may be useful

  • Typical parameters:

    • Population size

    • Parent selection operators

    • Offspring size

    • Survival selection

    • Mutation operators

    • Crossover operators



Fubos futility based offspring sizing
FuBOS: Futility-Based Offspring Sizing

  • Minimize wasted computation effort


Approach
Approach

  • Look at change in average fitness of the offspring

  • Average fitness of all n offspring

  • Average fitness of n-1 previously created offspring

  • Threshold value


Experimental setup
Experimental Setup

  • Compared FuBOS-EA and manually tuned EA (OOS-EA)

  • FuBOS-EA uses ε=.001

  • Test problems: DTRAP, SAT, and ONEMAX

  • Used population sizes of 100, 500, 1000

  • All tests used same parameters

  • Performance compared using One-Way ANOVA with significance level of .05







Conclusions
Conclusions

  • Competitive performance

  • Extra parameter


Fubos future work
FuBOS Future Work

  • The “epsilon problem”

  • Genetic Diversity

  • Parent Selection

  • Combine with dynamic population sizing



The importance of age
The Importance of Age

  • Age significantly impacts survival in natural populations


Methods

F

i

S

S

i

AGE

F

B

Methods

  • Survival chance (Si) of an individual is based on age and fitness

  • Main Equation

Fitness of i

=

Best Fitness


Survival chance from age

S

R

(

AGE

)

AGE

A

Survival Chance from Age

  • Age is tracked by individual, and is incremented every generation

  • Two equations explored for SAGE

  • Equation 1 (ABPS-EA1): linear decrease

1

=

-

Rate of decrease from age


Survival chance from age cont d

N

AG

S

AGE

Survival Chance from Age (cont’d)

  • Equation 2 (ABPS-EA2): more dynamic

Number of individuals in the same age group

AGE

1

=

-

-

2

P

2

G

Population size

Number of generations the EA will run


Survival chance from age cont d1

N

AGE

=

-

-

AG

S

1

AGE

2

P

2

G

Survival Chance from Age (cont’d)

  • Effects of

    • More individuals of the same age will decrease their survival chance

    • Age will decrease survival chance relative to the maximum age (G)

NAG

Si


Experimental setup1
Experimental Setup

  • Testing done on TSP (size 20/40/80)

  • Offspring size is constant

  • Compared to a manually tuned EA

  • Examine effects of

    • Initial population size

    • Offspring size

  • Tracked population statistics

    • Size

    • Average age

    • Global best fitness (GBF)


Performance results tsp size 20
Performance Results - TSP size 20

Average over 30 runs

Global best fitness

ABPS-EA1 -

ABPS-EA2 -


Performance results tsp size 40
Performance Results - TSP size 40

Average over 30 runs

Global best fitness

ABPS-EA1 -

ABPS-EA2 -




Equation with fitness scaling

F

F

F

i

i

W

S

S

AGE

AGE

F

F

F

B

B

W

Equation with Fitness Scaling

  • Attempt to fix the lack of selection pressure from fitness

  • New Main Equation

Fitness of i

-

Fitness Scaling

S

S

=

=

i

i

-

Best Fitness

Worst Fitness


Initial performance analysis from fitness scaling equation
Initial Performance Analysis from Fitness Scaling Equation

Average over 30 runs

using

Global best fitness


Initial performance analysis from fitness scaling equation cont d
Initial Performance Analysis from Fitness Scaling Equation (cont’d)

  • Independence from initial population size was maintained

  • Dynamic adjustment of population size during the run was improved

  • Additional selection pressure from elitism improved performance slightly


Abps conclusions
ABPS Conclusions (cont’d)

  • Independence from initial population value was achieved

  • Autonomous adjustment of population size during a single EA run was successful

  • Fitness scaling is needed for ABPS to work on more difficult problems


Abps future work
ABPS Future Work (cont’d)

  • Further exploration of fitness scaling methods

  • Test on other difficult problems

  • Compare to other dynamic population sizing schemes

  • Implement age-based offspring sizing


Impact

Impact (cont’d)


Impact1
Impact (cont’d)

  • Increases industry usability

  • Higher performance EAs

  • Progress towards completely parameterless EA


Questions

Questions? (cont’d)



Experimental setup2
Experimental Setup (cont’d)

  • DTRAP

  • SAT

  • ONEMAX


ad