Adaptive Clustering-Based Differential Evolution for Multimodal Optimization

Abstract

Multimodal optimization problems which widely exist in the scientific research and engineering applications, has aroused a wide interest of researchers. For solving multimodal optimization problems, numerous niching algorithms have been proposed to locate and track multiple optima. However, most of these algorithms need a very strict choice of the population size parameter. This paper presents a new niching differential evolution algorithm which adaptively adjusts population size during the evolution. Particularly, we propose three techniques for performance enhancement: a heuristic clustering method, a population adaptation strategy, and an auxiliary movement strategy for the best individuals. The algorithm divides the population into several subpopulations and adaptively adjust the number of individuals and subpopulations according to the evolutionary state. In this way, the diversity of population is increased, while the computational cost is reduced. Experimental results verify that the proposed algorithm outperforms the other niching algorithms for multimodal optimization.