Effects of hierarchical migration in a parallel distributed parameter-free GA

Effects of hierarchical migration methods in a parallel distributed Parameter-free GA (PfGA) are described. The PfGA is a compact and robust algorithm that extracts a local population from whole population, and evolve it by adaptively changing the size of subpopulation. We propose hierarchically parallel distributed architectures with migration methods for the PfGA implementing them in a parallel machine. In evaluating many function optimization problems as recent benchmark tests, we verified that the proposed parallel PfGA architectures with the migration methods effectively decrease the number of evaluations to converge with the success rates held or improved by increasing the number of subpopulations.