Adapting genotype-phenotype-mapping by using redundant real representation

A concept of adapting genotype-phenotype-mapping is proposed. The concept is that a neighborhood in a genotype space that is mapped into separated areas around global and local optima in a phenotype one is evolutionally obtained. Two things are required to realize an optimization method according to the concept. One is a genotype-phenotype-mapping, which can map a neighborhood in a genotype space into several areas in the phenotype one. Another is a search method, which can preserve search areas in the genotype space as structures. The genotype-phenotype-mapping and the search method that satisfy with those requirements are proposed. The numerical optimization method combining those proposed methods is applied to artificial simple test functions, and it is shown that the method realizes the concept of adapting genotype-phenotype-mapping.