Topics in Evolutionary Computation and Model-Based Clustering

Abstract

Topics in evolutionary computation and mixture model-based clustering are explored, as well as some work at the intersection of the two. First, complex fitness landscapes and evolutionary computation techniques are explored using cellular automata, and single parent techniques are developed to improve evolutionary techniques for locating and generalizing automata rules. Next, mixture model-based approaches to clustering are considered. Parameter estimation in mixture model-based approaches to statistical learning is notoriously difficult, and an evolutionary computation approach is proposed and illustrated on the well-established Gaussian mixture model. Next, a mixture of variance-gamma factor analyzers model is developed to facilitate the flexible clustering of high-dimensional data. This thesis concludes with a discussion and suggestions for future work.