Masters Theses

Keywords and Phrases

Linkage Tree Genetic Algorithm; Self-configuring crossover

Abstract

"Evolutionary Algorithms (EAs) have shown great potential to solve complex real world problems, but their dependence on problem specific configuration in order to obtain high quality performance prevents EAs from achieving widespread use. While it is widely accepted that statically configuring an EA is already a complex problem, dynamic configuration of an EA is a combinatorially harder problem. Evidence provided here supports the claim that EAs achieve the best results when using dynamic configurations. By designing methods that automatically configure parts of an EA or by changing how EAs work to avoid configurable aspects, EAs can be made more robust, allowing them better performance on a wider variety of problems with less requirements on the user.

Two methods are presented in this thesis to increase the robustness of EAs. The first is a novel algorithm designed to automatically configure and dynamically update the recombination method which is used by the EA to exploit known information to create new solutions. The techniques used by this algorithm can likely be applied to other aspects of an EA in the future, leading to even more robust EAs. The second is an existing set of algorithms which only require a single configurable parameter. The analysis of the existing set led to the creation of a new variation, as well as a better understanding of how these algorithms work. Both methods are able to outperform more traditional EAs while also making both easier to apply to new problems. By building upon these methods, and perhaps combining them, EAs can become even more robust and become more widely used"--Abstract, page iv.

Advisor(s)

Tauritz, Daniel R.

Committee Member(s)

Erçal, Fikret
Wunsch, Donald C.

Department(s)

Computer Science

Degree Name

M.S. in Computer Science

Publisher

Missouri University of Science and Technology

Publication Date

Spring 2012

Journal article titles appearing in thesis/dissertation

  • Meta-evolved empirical evidence of the effectiveness of dynamic parameters
  • Self-configuring crossover
  • Linkage tree genetic algorithms: variants and analysis

Pagination

x, 59 pages

Rights

© 2012 Brian Wesley Goldman, All rights reserved.

Creative Commons Licensing

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 License.

Document Type

Thesis - Open Access

File Type

text

Language

English

Subject Headings

Evolutionary programming (Computer science)Liaison theory (Mathematics)

Thesis Number

T 9964

Print OCLC #

815957035

Electronic OCLC #

786169107

Download

Share

 
COinS