Publications

Abstract : Distributed Differential Evolution, advancement in Differential Evolution (DE) algorithm, is based on the principle of cooperation and co-evolution. It provides multiple search space perspectives, alternate search paths and a more balanced exploitation and exploration capabilities to DE. Altogether it improves solution quality and prevents premature convergence and stagnation to a greater extent, compared to serial DE. The solution quality of Distributed DE is based on the choice of migration parameters. One of the most influencing migration parameter is the topology used. The aim of this work is to empirically analyze the performance of Distributed DE for varying migration topologies. Migration topologies differ from each other on the basis of degree and interconnectivity of nodes. This paper empirically analyzes the performance difference of distributed differential evolution algorithm with varying migration topologies, on a set of benchmarking problems. The migration topologies used in our experiments are basic ring and its variants, star, cartwheel, torus and mesh. Experimental results have shown that no single topology can said to be good for all optimization problems. It depends on the complexity and type of objective function to be optimized. Experimental analyses have also exposed the influence of DE variant, employed in different islands, on the performance ordering of topologies. The influence of other parameters like the selection policy and replacement policy are also found to be crucial. © Research India Publications.