Evolutionary many objective optimization based on bidirectional decomposition

作     者：LYU Chengzhong LI Weimin 

作者机构：Air and Missile Defense CollegeAir Force Engineering University 

出 版 物：《Journal of Systems Engineering and Electronics》 (系统工程与电子技术（英文版）)

年 卷 期：2019年第30卷第2期

页      面：319-326页

核心收录：

学科分类：07[理学] 08[工学] 

主　　题：many objective optimization bidirectional decomposition reference update evolutionary algorithm 

摘      要：The decomposition based approach decomposes a multi-objective problem into a series of single objective subproblems, which are optimized along contours towards the ideal point. But non-dominated solutions cannot spread uniformly, since the Pareto front shows different features, such as concave and convex. To improve the distribution uniformity of non-dominated solutions, a bidirectional decomposition based approach that constructs two search directions is proposed to provide a uniform distribution no matter what features problems have. Since two populations along two search directions show differently on diversity and convergence, an adaptive neighborhood selection approach is presented to choose suitable parents for the offspring generation. In order to avoid the problem of the shrinking search region caused by the close distance of the ideal and nadir points, a reference point update approach is presented. The performance of the proposed algorithm is validated with four state-of-the-art algorithms. Experimental results demonstrate the superiority of the proposed algorithm on all considered test problems.