Differential evolution in high-dimensional search spaces

Abstract

A possible way of dealing with a high dimensional problem space is to divide it up into smaller parts, and to have each part optimized by a separate population. A mechanism is then defined to construct a complete solution from the subpopulations, and to evaluate the entities contained in the subpopulations. This form of cooperation has been successfully applied to particle swarm optimization (PSO), by [1] in the cooperative split PSO, and to genetic algorithms, in the cooperative coevolutionary genetic algorithm, developed by [2], on which the cooperative split PSO is based. This paper investigates cooperation in differential evolution (DE) with the aim of determining the effects of multiple participants in dealing with high-dimensional problem spaces.