Performance Study of a Multi-Deme Parallel Genetic Algorithm with Adaptive Mutation

This paper presents a performance study of a parallel, coarse-grained, multiple-deme Genetic Algorithm (GA) with adaptive mutation. The effect of varying migration period and number of subpopulations upon the GA is evaluated. Using common unimodal and multimodal objective functions, this study measures the convergence velocity and solution quality for the proposed genetic algorithm. In this paper, we briefly survey previous work in static and adaptive control parameters and parallel genetic algorithms (PGAs). Experimental results show that migration period and the number of subpopulations significantly influence the performance of the genetic algorithm.