确定性排挤小生态技术的遗传漂移分析  

Genetic drift analysis of the niching method with deterministic crowding

作　　者：余晓峰[1,2] 谭竹梅[1,2] 郭观七[1,2] 

机构地区：[1]湖南理工学院机械与电气工程系,湖南岳阳414000 [2]湖南理工学院计算机与信息工程系,湖南岳阳414006

出　　处：《湖南理工学院学报（自然科学版）》2004年第3期34-38,共5页Journal of Hunan Institute of Science and Technology(Natural Sciences)

基　　金：国家自然科学基金项目 :编号 5 0 2 75 170;湖南省教育厅科学研究基金资助项目 :编号 2 0 0 1C3 80 ;2 0 0 2A0 5 2

摘　　要：介绍多峰搜索空间的等价类模型 ,应用该模型分析确定性排挤 (DC)小生态遗传算法的遗传漂移现象。分析结果表明 ,DC的替换错误导致类形成与类维持的矛盾 ,DC不能克服类之间的遗传漂移 ,且倾向于维持高适应值的类而丢失低适应值的类。因此 ,对于复杂多峰优化问题 ,DC不能在单一种群中并行地维持多个全局或局部最优解。Equivalence class model of multimodal search space is introduced. It is used to analyze the genetic drift in niching genetic algorithms using deterministic crowding method (DC). The results show that the replacement errors of similar individuals in DC cause the confliction between generating new classes and maintaining the existing classes, and DC is not able to remove the genetic drift in evolutionary computation, but inclined to maintain the classes with higher fitness and lose the ones with lower fitness. Thus DC is not capable of finding and maintaining multiple global or local optima in parallel in a single population when used to optimize complicated multimodal problems.

关 键 词：进化算法 小生态 确定性排挤 遗传漂移 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]