Abstract

A new genetic algorithm which uses a 3-parent uniform crossover operator is developed and analyzed. Uniform crossover operators are shown to be based on the premise that all bit-level genetic information should be passed from parents to children. The 3-parent uniform crossover operator is shown to adhere to this premise. The 3-parent uniform crossover operator is shown to be better than the 2-parent uniform crossover operator on the De Jong test functions.

Two new genetic algorithms which use 3-parent traditional crossover operators are developed and analyzed. The first uses a strategy of randomly selecting 3 of the 6 children resulting from 3-parent reproduction. The second uses a strategy of selecting the best 3 of the 6 children resulting from 3-parent reproduction. Each of the 3-parent traditional crossover operators is shown to be superior to the 2-parent traditional crossover operator on the De Jong test functions. The strategy of selecting the best 3 out of 6 children is shown to be superior to the strategy of randomly selecting 3 out of 6 children.

In addition to these 3-parent genetic algorithms, a relationship between the Metropolis algorithm from simulated annealing and the two-membered evolution strategy is developed. The Metropolis algorithm is shown to be a special case of the two- membered evolution strategy.

Department(s)

Computer Science

Comments

This report is substantially the Ph.D. dissertation of the first author, completed July 1993.

Report Number

CSC-93-22

Document Type

Technical Report

Document Version

Final Version

File Type

text

Language(s)

English

Rights

© 1993 University of Missouri--Rolla, All rights reserved.

Publication Date

01 Jul 1993

Share

 
COinS