History archive assisted niching differential evolution with variable neighborhood for multimodal optimization

Abstract

Multimodal optimization problems (MMOPs) require the algorithms to find multiple global or local optima in a single run, which is considered a difficult task. In recent years, niching techniques have received widespread attention from researchers and extensively recognized as efficient methods for MMOPs. However, most niching techniques have the disadvantage of parameters settings. Moreover, the information of historical individuals is not fully used for evolution. In this paper, a history archive assisted niching differential evolution with variable neighborhood (HANDE/VN) is proposed. The main features are as follows: (i) a variable neighborhood strategy dynamically controls the neighborhood size of individual during the evolution, which can balance exploration and exploitation to some extent; (ii) a new mutation operation based on history archive can fully exchange the information between history archive and current population to enhance the search ability and help the algorithm locate multiple global solutions more accurately; (iii) a local optima processing based on history archive deal with the individuals trapped in the local optima. To demonstrate the performance of HANDE/VN, 20 MMOPs from CEC2013 were selected as test suite. Compared with state-of-the-art multimodal algorithms, the proposed approach obtains better results in terms of peak ratio and success rate.