The Nature of Mutation in Genetic Algorithms
Robert Hinterding, Harry Gielewski, Thomas Peachey · 1995
Normally in Genetic Algorithms, mutation is considered a background operator and the genes are considered to be the binary bits of the chromosome. In this paper we take a different viewpoint. We treat the variables of numerical functions as the genes, and consider the mutation of these genes. We also investigate the role of mutation as an independent reproduction operator. Our results show the value of this view, and explain some previous comparisons with Evolutionary Strategies. 1 Introduction In this paper we look at the nature of mutation in Genetic Algorithms(GAs) used in optimisation of numerical functions. Current theory of GAs considers that mutation is a background operator and just used to provide bits lost by crossover (Holland, 1992). This relies on the view that the genes in a chromosome are binary bits, and is supported via the current schema theory (Holland, 1992; Goldberg, 1989). Further current guidelines recommend using genes with a small alphabet so that the largest ...