计算机科学与技术

基于分方向选择搜索的多目标进化算法

展开
  • 1.华南理工大学 计算机科学与工程学院,广东 广州 510006;2.广东工业大学 计算机学院,广东 广州 510090;3.华南理工大学 软件学院,广东 广州 510006
李学强(1983-),男,博士生,主要从事多目标优化研究

收稿日期: 2010-05-21

  网络出版日期: 2011-01-02

基金资助

国家自然科学基金资助项目(60873078,61003066,61070033);高等学校博士学科点专项科研基金资助项目(20090172120035);广东省自然科学基金资助项目(9251009001000005);广东省科技计划项目(2010B0504(0)011,2010B080701070,2008B080701005);华南理工大学中央高校基本科研业务费专项资金资助项目(2009ZM0052)

Multi-Objective Evolutionary Algorithm Based on Direction Selection Search

Expand
  • 1. South China university of technology, computer science and engineering college, guangdong guangzhou 510006; 2.Computer college of guangdong university of technology, guangdong guangzhou 510090; 3. South China university of technology of software institute, guangdong guangzhou 510006
李学强(1983-),男,博士生,主要从事多目标优化研究

Received date: 2010-05-21

  Online published: 2011-01-02

Supported by

国家自然科学基金资助项目(60873078,61003066,61070033);高等学校博士学科点专项科研基金资助项目(20090172120035);广东省自然科学基金资助项目(9251009001000005);广东省科技计划项目(2010B0504(0)011,2010B080701070,2008B080701005);华南理工大学中央高校基本科研业务费专项资金资助项目(2009ZM0052)

摘要

Pareto解空间为分区域连续时,基于极大极小策略和确定权重的多目标进化算法在进化的每一代都会在无最优解的区域搜索解点.为解决此问题,文中提出了一种新的判断机制,根据每个权重在不同方向上选取的解点集来判断权重对应的Pareto有效解区域是否存在最优解点,并以此来调整搜索的区域;为了避免算法陷入局部收敛和种群早熟,提出了分方向选择搜索的方法,使用了外部存储的机制.采用文中算法对常见的测试函数和CEC2009多目标竞赛中较难的测试函数进行了测试,结果表明该算法是有效的.

本文引用格式

李学强 郝志峰 黄翰 . 基于分方向选择搜索的多目标进化算法[J]. 华南理工大学学报(自然科学版), 2011 , 39(2) : 130 -135 . DOI: 10.3969/j.issn.1000-565X.2011.02.022

Abstract

When the Pareto solution space is piecewise continuous,the multi-objective optimization algorithm based on the min-max strategy and the pre-set weights may result in the search of the best value by each generation in non-optimal area.In order to solve this problem,a novel judge mechanism is proposed to estimate whether there exists an optimal solution in the Pareto solution space according to the solution set selected in different directions,and the search area is thus adjusted.Moreover,in order to avoid local convergence and prematurity of the proposed algorithm,a direction selection search method based on the external storage mechanism is put forward.Finally,the proposed algorithm is used to test some common multi-objective test functions and some difficult multi-objective test functions in CEC2009 Race.The results indicate that the algorithm is effective.

文章导航

/