An evolutionary algorithm based on decomposition for multimodal optimization problems

This paper presents a non-parameter method to identify the peaks of the multi-modal optimization problems provided that the peaks are characterized by a smaller objective values than their neighbors and by a relatively large distance from points with smaller objective value. Using the identified peaks as the seeds, we decompose the population into some sub-populations and dynamically allocate the computational effort to different subpopulations. We evaluate the proposed approach on the CEC2015 single objective multi-niche optimization problems. The promising experimental results show its efficacy.