Competition-Based Distributed Differential Evolution

Abstract

Differential evolution (DE) is a simple and efficient evolutionary algorithm for global optimization. In distributed differential evolution (DDE), the population is divided into several sub-populations and each sub-population evolves independently for enhancing algorithmic performance. Through sharing elite individuals between sub-populations, effective information is spread. However, the information exchanged through individuals is still too limited. To address this issue, a competition-based strategy is proposed in this paper to achieve comprehensive interaction between sub-populations. Two operators named opposition-invasion and cross-invasion are designed to realize the invasion from good performing sub-populations to bad performing subpopulations. By utilizing opposite invading sub-population, the search efficiency at promising regions is improved by opposition-invasion. In cross-invasion, information from both invading and invaded sub-populations is combined and population diversity is maintained. Moreover, the proposed algorithm is implemented in a parallel master-slave manner. Extensive experiments are conducted on 15 widely used large-scale benchmark functions. Experimental results demonstrate that the proposed competition-based DDE (DDE-CB) could achieve competitive or even better performance compared with several state-of-the-art DDE algorithms. The effect of proposed competition-based strategy cooperation with well-known DDE variants is also verified.