摘要
多目标进化算法应用非常广泛,但易陷入局部Pareto前沿.为了提高多目标进化算法平衡全局探索与局部开发的能力,使算法收敛到完整的Pareto前沿,本文提出采用个体进化状态判定策略的分解类多目标进化算法(MOEA/D_PE),MOEA/D_PE算法采用个体进化状态判定策略,判定个体当前的进化状态,然后为个体选择适合其进化状态的变异算子,从而提高算法平衡全局探索与局部开发的能力,使算法收敛到完整的Pareto前沿.实验研究表明,MOEA/D_PE算法在测试函数中表现出,比非支配排序多目标遗传算法和分解类多目标进化算法更好的收敛性和多样性,能够更好地收敛到完整的Pareto前沿.
Multi-objective evolutionary algorithm is widely used but easily trapped in local Pareto front. In order to improve the ability of multi-objective evolutionary algorithm balancing the global exploration and local development, and to make the algorithm converge to the complete Pareto front, in this paper, a multi-objective evolutionary algorithm of decomposition with individual evolutionary state judging strategy (MOEA/D_PE) is proposed. The MOEA/D_PE algorithm uses the individual evolutionary state judging strategy to judging the current evolutionary state of the individual,Then, an appropriate mutation operator corresponding to its evolutionary state is selected for the individual to improve the ability of the algorithm balancing the global exploration and local development, and to make algorithm converge to the complete Pareto front. The experimental results show that the MOEA/D_PE algorithm shows better conver- gence and diversity in the test function than the existing non-dominated sorting multi-objective genetic algorithm and multi-objective evolutionary algorithm of decomposition, and can converge to the complete Pareto front.
作者
李浩君
刘中锋
王万良
张征
张鹏威
LI Hao-jun;LIU Zhong-feng;WANG Wan-liang;ZHANG Zheng;ZHANG Peng-wei(College of Education Science and Technology,Zhejiang University of Technology,Hangzhou 310023,China;College of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310023,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2018年第8期1668-1673,共6页
Journal of Chinese Computer Systems
基金
国家社科基金年度项目(16BTQ084)资助
关键词
个体进化状态
变异算子
多目标
分解
individual evolutionary state
mutation operator
multi-objective
decomposition