This work examines the impact that genes with activation relationships have on Genetic Algorithms (GAs). These activation relationships allow genes to control whether other associated genes get expressed in the phenotype. More specifically, this thesis investigates the effect that the incorporation of control levels (tiers of genes which determine the activation of lower genes) have on GAs which are used to generate feed forward Artificial Neural Networks (ANNs). In order to evaluate the performance of different levelled Structured Genetic Algorithms (SGAs; they are GAs which possess control levels), numerous experiments were conducted, utilising the four input XOR, Mackey-Glass and Breast Cancer data sets. In addition, the thesis derives four mathematical models which describe how SGA redundancy changes as more control levels are incorporated. This thesis also presented and implemented a novel four level Structurally Evolved Neural Network Algorithm (SENNGA). Furthermore, it demonstrated the novel implementation of a three level SENNGA. Empirical results show that the incorporation of control levels can improve convergence speed and accuracy, up until the optimum number of control levels is reached. In all experiments increasing the number of control levels had the effect of encouraging the production of leaner and more efficient ANNs. Three and four level SENNGAs can demonstrate superior performance in terms of reducing the training error and generalisation. However this does not apply in all cases. Of particular note, three and four level SENNGAs have exhibited a reduced level of generalisation in the breast cancer experiments, compared two level ones. / Doctor of Philosophy (PhD)
| Identifer | oai:union.ndltd.org:ADTP/181834 |
| Date | January 2006 |
| Creators | Molfetas, Angelos, University of Western Sydney, College of Health and Science, School of Computing and Mathematics |
| Source Sets | Australiasian Digital Theses Program |
| Language | English |
| Detected Language | English |
Page generated in 0.0034 seconds