An adaptive neighborhood-based speciation differential evolution for multimodal optimization

Abstract

Multimodal optimization problems (MMOPs) are single-objective optimization problems with multiple optimal solutions. It is essential to find as many optimal solutions as possible to make a robust decision. However, it is a very challenging task. In this paper, an enhanced adaptive neighborhood-based speciation differential evolution (EANSDE) is proposed to solve the MMOPs effectively, which can be featured as: (i) The parameters are adaptively controlled to alleviate the fine-tuning of parameters by the users. (ii) An external archive is introduced to store the inferior solutions discarded in each generation. When the population size is reduced to a preset threshold, the solutions in the archive are merged with the current population in the following search. (iii) The crowding relieving mechanism is proposed to remove extremely similar individuals from population. Compared with related methods, experimental results indicate the superiority of EANSDE on the 20 benchmark MMOPs in CEC-2013.