基于动态拥挤距离的自适应变异演化算法  

ADAPTIVE MUTATION EVOLUTIONARY ALGORITHM BASED ON DYNAMIC CROWDING DISTANCE

作　　者：杨林根[1] 

机构地区：[1]广东白云学院计算机系

出　　处：《计算机应用与软件》2012年第1期280-283,共4页Computer Applications and Software

摘　　要：多目标演化算法的研究目标是使算法种群快速收敛并均匀分布于问题的非劣最优域。根据个体的非支配排序级数设计了一种自适应变异步长的柯西变异算子,对变异越界处理进行了改进;并定义和使用动态拥挤距离来保持群体中个体的均匀分布。最后通过对测试函数的实验,验证了算法的可行性和有效性。The goal of the research on multi-objective evolutionary algorithm is to make the algorithm population fast converge and uniformly distributed in Pareto optimal domain of the issue.In this paper we design an adaptive mutation stepsize of Cauchy mutation operator based on non-dominated sorting series of the individual,and improve the treatment on variation of cross-border;and define and use dynamic crowding distance to maintain the uniform distribution of individuals.At last we prove the feasibility and validity of the new approach through experiment on test functions.

关 键 词：演化算法 多目标优化 自适应变异 动态拥挤距离 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]