Neutrality through transcription and translation in genetic algorithm representation

This paper examines the use of the biological concepts of transcription and translation, to introduce neutrality into the representation of a genetic algorithm (GA). The aim of the paper is to attempt to identify problem characteristics which may benefit from the inclusion of neutrality, through a basic adaptation of the concepts of transcription and translation, to create a genotype-phenotype map (GP-map) which introduces phenotypic variability. Neutrality can be viewed as a situation where a number of different genotypes represent the same phenotype. A modification of De Jong's classic test suite was used to compare the performance of a simple generic algorithm (SGA) and a multi layered mapping genetic algorithm (MMGA), which incorporates the concepts of transcription and translation into its GP-map. The modified De Jong test suite was chosen as it is well understood and has been used in numerous comparisons over the years, thus allowing us to contrast the performance of the MMGA against other GA variations as well as attempting to identify problem characteristics in isolation. Initial results indicate that the neutrality introduced through the multi-layered mapping can prove beneficial for problems containing certain characteristics, in particular multidimensional, multimodal, continuous and deterministic.