Performance Comparisons of Genotype to Phenotype Mapping Schemes for Diploid Representations in Changing Environments

When working in changing environments, it is important for a solution approach to adapt to the change and follow the new optima. For genetic algorithms, using diploid representations is one of the methods for addressing performance and diversity issues in such environments. When using a diploid representation for individuals, the choice of a good genotype to phenotype mapping mechanism is a very important factor. The main aim of this study is to give performance comparisons of some of the existing genotype to phenotype mapping approaches, show their weaknesses and propose a new, adaptive approach that will perform better in the cases where these fail. The proposed approach is compared against the Ng-Wong and the additive diploidy approaches with a dominance change mechanism, using a controlled 0/1 Knapsack problem. All three approaches tested for this study, show similar performances for oscillating change. In the discussed random change case, the proposed approach shows better performance in automatically adapting to the change.