A Markov Process Approach to Redundancy in Genetic Algorithms

Abstract

We propose and investigate two Markov chain models in order to provide recommendations on the selection of genotype redundant encoding schemes in genetic algorithms (GA). It was deduced from these Markov chains why classic encoding methods lead to a insufficiently broad search of GA’s. Then, modifications, concerning additional redundancy in genotype encoding algorithms, are proposed and investigated regarding its influence on mutations and recombinations. Finally, the results of extensive simulation studies are reported that indicate classes of goal functions for which these modifications provide better search results and/or smaller variances of them.