Doing Genetic Algorithms the Genetic Programming Way

Abstract

This paper describes the GAuGE system, GeneticAlgorithms using Grammatical Evolution, which usesGrammatical Evolution to perform as a positionindependent Genetic Algorithm. Gauge has already beensuccessfully applied to domains such as bit level,sorting and regression problems, and our experiencesuggests that it evolves individuals with a similardynamic to Genetic Programming. That is, there is ahierarchy of dependency within the individual, and, asevolution progresses, those parts at the top of thehierarchy become fixed across a population. We look atthe manner in which the population evolves therepresentation at the same time as optimising theproblem, and demonstrate there is a definite emergenceof representation.