Differential evolution with dynamic neighbourhood learning strategy-based mutation operators

As the core operator of differential evolution (DE), mutation is crucial for guiding the search. However, in most DE algorithms, the parents in the mutation operator are randomly selected from the current population, which may lead to DE being slow to exploit solutions when facing complex problems. In this study, a dynamic neighbourhood learning (DNL) strategy is proposed for DE to alleviate this drawback. The new proposed DE framework is named DE with DNL-based mutation operators (DNL-DE). Unlike the original DE algorithms, DNL-DE uses DNL to dynamically construct neighbourhood for each individual during the evolutionary process and intelligently select parents for mutation from the defined neighbourhood. In this way, the neighbourhood information can be effectively utilised to improve the performance of DE. Furthermore, two instantiations of DNL-DE with different parent selection methods are presented. To evaluate the effectiveness of the proposed algorithm, DNL-DE is applied to the original DE algorithms, as well as several advanced DE variants. The experimental results demonstrate the high performance of DNL-DE when compared with other DE algorithms.