Performance-driven adaptive differential evolution with neighborhood topology for numerical optimization

This paper presents a novel differential evolution algorithm for numerical optimization by making full use of the neighborhood information to balance exploration and exploitation. To effectively meet the search requirement of each individual, a neighborhood-based adaptive mutation strategy is developed by using the ring topology to construct an elite individual set and adaptively choosing a suitable elite individual to guide its search according to its neighborhood performance. Then, a neighborhood-based adaptive parameter setting is designed to improve the suitability of parameters for each individual by utilizing the feedback information of population and its neighbors simultaneously. Furthermore, a restart mechanism is proposed to further enhance the performance of algorithm by adaptively strengthening the search abilities of unpromising individuals, removing the worse individuals and randomly replacing some individuals with Gaussian Walks. Differing from the existing DE variants, the proposed algorithm adaptively guides the search and suitably adjusts the parameters for each individual by using its neighborhood performance, and strengthens the exploitation and exploration by removing the worse individuals and randomly replacing some individuals. Then it could properly adjust the search ability of each individual, and effectively balance diversity and convergence. Compared with 16 typical algorithms, the numerical results on 30 IEEE CEC2014 benchmark functions show that the proposed algorithm has better performance. (C) 2019 Elsevier B.V. All rights reserved.