Title

Adaptive Selection Operators in Evolutionary Algorithms

Presenter Information

Nathaniel Kamrath

Department

Computer Science

Major

Computer Science

Research Advisor

Tauritz, Daniel R.

Advisor's Department

Computer Science

Funding Source

Opportunities for Undergraduate Research (OURE)

Abstract

Selection is a core genetic operator in many evolutionary algorithms (EAs). The performance of EAs on a given problem is dependent on properly configuring selection. A small set of common selection operators is used in the vast majority of EAs, typically fixed for the entire evolutionary run. Selecting which crossover operator to use and tuning its associated parameters to obtain acceptable performance on a specific problem is often a time consuming, manual process. Even then a custom selection operator may be required to achieve optimal performance. Finally, the best crossover configuration may be dependent on the state of the evolutionary run.

This work introduces Adaptive Selection Operators which address these shortcomings while relieving the user from the burden of selection operator configuration. Results are presented showing it to outperform the traditional selection operators k-tournament, truncation, and fitness proportionate selection on the Rosenbrock benchmark problem.

Biography

Nathaniel is currently a senior in Computer Science and an Undergraduate Research Assistant in the Natural Computation Laboratory.

Research Category

Sciences

Presentation Type

Poster Presentation

Document Type

Poster

Award

Sciences poster session, Second place

Location

Upper Atrium/Hall

Presentation Date

16 Apr 2014, 9:00 am - 11:45 am

This document is currently not available here.

Share

COinS
 
Apr 16th, 9:00 AM Apr 16th, 11:45 AM

Adaptive Selection Operators in Evolutionary Algorithms

Upper Atrium/Hall

Selection is a core genetic operator in many evolutionary algorithms (EAs). The performance of EAs on a given problem is dependent on properly configuring selection. A small set of common selection operators is used in the vast majority of EAs, typically fixed for the entire evolutionary run. Selecting which crossover operator to use and tuning its associated parameters to obtain acceptable performance on a specific problem is often a time consuming, manual process. Even then a custom selection operator may be required to achieve optimal performance. Finally, the best crossover configuration may be dependent on the state of the evolutionary run.

This work introduces Adaptive Selection Operators which address these shortcomings while relieving the user from the burden of selection operator configuration. Results are presented showing it to outperform the traditional selection operators k-tournament, truncation, and fitness proportionate selection on the Rosenbrock benchmark problem.