多目标优化差分进化算法  

Differential Evolution Algorithm for Multi-Objective Optimization

作　　者：敖友云[1] 迟洪钦[2] 

机构地区：[1]安庆师范学院计算机与信息学院,安徽安庆246011 [2]上海师范大学计算机系,上海200234

出　　处：《计算机工程与科学》2011年第9期88-94,共7页Computer Engineering & Science

摘　　要：个体的适应度赋值和群体的多样性维护是进化算法的两个关键问题。首先,一方面,定义了Paretoε-支配关系的相关概念,通过Paretoε-支配关系确定个体的强度Pareto值,根据个体的强度Pareto值对群体进行Pareto分级排序,实现优胜劣汰;另一方面,使用拥挤距离估算个体的拥挤密度,淘汰位于拥挤区的一些个体,维持群体的多样性。然后,根据差分进化算法的特点,使用适当的进化策略和控制参数,给出了一种用于求解多目标优化问题的差分进化算法DEAMO。最后,数值实验表明,DEAMO在求解标准的多目标优化问题时性能表现优良。Fitness assignment of individuals and diversity maintenance of population are two key techniques of evolutionary algorithms. First, on the one hand, this paper introduces some related concepts of Pareto ε-dominance which can determine the strength Pareto values of the individuals of population, according to the strength Pareto values of individuals, some better individuals are selected into the offspring population by the technique of Pareto ranking; on the other hand, in order to maintain the diversity of population, a crowded-density method is introduced to remove some individuals that are located in the crowed regions. Then, according to some characteristics of differential evolution （DE）, through using the appropriate DE strategies and control parameters, this paper proposes a differential evolution algorithm for multi-objective optimization, which is called DEAMO. Fi nally, numerical experiments show that DEAMO can perform well when tested on several benchmark multi-ob- jective optimization problems.

关 键 词：多目标优化 差分进化 进化算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]