改进的自适应邻域的多目标进化算法  

Improved multi-objective evolutionary algorithm based on adaptive neighborhood

作　　者：薛胜军[1] 杨明[1] 

机构地区：[1]南京信息工程大学计算机与软件学院,南京210044

出　　处：《计算机工程与应用》2011年第25期49-53,共5页Computer Engineering and Applications

基　　金：国家自然科学基金(No.60702076);上海市信息安全综合管理技术研究重点实验室开放课题(No.AGK2009001)~~

摘　　要：提出了一种新的自适应邻域的多目标进化算法,该算法采用自适应邻域的方法维护群体的分布性。探讨了根据当前群体情况进行自适应改变邻域半径,避免了传统邻域策略所引起的邻域半径的取值影响群体分布性的问题。另外,利用自适应邻域半径和拥挤距离进行密度估计,使密度小的个体得到保留。实验结果表明,所讨论的方法是有效的,在保持群体分布性上优于NSGAII和NMOEA。A novel Multi-Objective Evolutionary Algorithm based on Adaptive Neighborhood（ANMOEA） is proposed.In the ANMOEA,an adaptive neighborhood method is used to maintain the diversity of the population.Moreover,this paper discusses that the radius of the neighborhood is adaptively changed by the situation of the current population,which avoids the problem that radius value of the neighborhood affects the diversity of the population in the traditional neighborhood strategy.In addition,adaptive neighborhood radius and crowding distance are applied to evaluate density of individuals,which preserves the small density of the individuals.The experimental results indicate that the discussed method is effective in maintaining the diversity of the population,which is significantly stronger than NSGAII and NMOEA multi-objective algorithm.

关 键 词：自适应邻域半径 多目标进化算法 邻域集 自适应邻域 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]