Main content

Analysing the Effectiveness of Different Parameter Optimization Techniques for Complex Genetic Algorithmsd

Show full item record

Title: Analysing the Effectiveness of Different Parameter Optimization Techniques for Complex Genetic Algorithmsd
Author: Kielstra, Immanuel Thomas
Department: Department of Mathematics and Statistics
Program: Mathematics and Statistics
Advisor: Ashlock, Daniel
Abstract: This thesis attempts to determine whether a modified version of the genetic algorithm developed by Ashlock et al. (2014) could be used to create graphs that are hard to colour for graph colouring algorithms, specifically the random sort (RS) method. In order to determine whether or not this is possible, an optimal parameter setting for the genetic algorithm must be obtained. The appropriate population size, iteration number and mutation number was determined using a uniform chromosomal regulatory sequence. This thesis then analysed the edit commands of the chromosomal regulatory sequence to determine general characteristics that could be used to optimize the chromosomal regulatory sequence. Modified versions of particle swarm optimization and differential evolution were also used to try to optimize the chromosomal regulatory sequence. This thesis shows that even though the edit commands do have some general characteristics, it is ineffective to use these characteristics to optimize the genetic algorithm. Both particle swarm optimization and differential evolution are effective at optimizing the genetic algorithm given appropriate conditions. The genetic algorithm appears to be effective at finding hard to colour graphs for the RS method. This genetic algorithm could be used to create graphs for other graph colouring algorithms. This thesis suggests modifications that could be made to the genetic algorithm based on some of the edit commands' general characteristics.
URI: http://hdl.handle.net/10214/14694
Date: 2018-12
Terms of Use: All items in the Atrium are protected by copyright with all rights reserved unless otherwise indicated.


Files in this item

Files Size Format View
Kielstra_Thomas_201812_Msc.pdf 3.407Mb PDF View/Open

This item appears in the following Collection(s)

Show full item record