Doing Genetic Algorithms the Genetic Programming Way

This paper describes the GAuGE system, Genetic Algorithms using Grammatical Evolution, which uses Grammatical Evolution to perform as a position independent Genetic Algorithm. Gauge has already been successfully applied to domains such as bit level, sorting and regression problems, and our experience suggests that it evolves individuals with a similar dynamic to Genetic Programming. That is, there is a hierarchy of dependency within the individual, and, as evolution progresses, those parts at the top of the hierarchy become fixed across a population. We look at the manner in which the population evolves the representation at the same time as optimising the problem, and demonstrate there is a definite emergence of representation.