Session:

LBP - Late Breaking Papers

Title:

Determining the Best Parent Selection Method for a Genetic Algorithm through Varying Problem Sizes and Complexities

Authors:

Daman Oberoi
Bart Ryler

Abstract:

We conduct an experiment to investigate which of three parent selection methods (elitism, fitness proportionate, and tournament) generates the greatest fitness in the fewest number of generations using a genetic algorithm (GA). The parent selection methods were applied to the problems of Maximum Ones, 3-Processor Scheduling, and Sorting, while in each case the problem size varied from 4 to 22. We show that for nearly all problem sizes and types, Tournament selection produces the best results.