摘要
基于群体搜索的演化算法求解多目标优化问题有独特的优势 ,多目标演化算法已有的研究大多为算法的设计和数值试验效果的比较 ,理论研究往往被忽视 .该文讨论了多目标演化算法的收敛性问题 ,针对一种网格化的简单易于实现的多目标演化算法模型定义了多目标演化算法强收敛和弱收敛等概念 ,给出了判断算法收敛性的一般性条件 ;在变异算子为高斯变异、目标函数连续的条件下 ,证明了提出的算法强收敛 .数值实验验证了算法的可行性和有效性 .
Evolutionary algorithms are especially suited for multi-objective optimization problems. Many evolutionary algorithms have been successfully applied to various multi-objective optimization problems, however, theoretical results on multi-objective evolutionary algorithms are scarce. This paper analyzes the convergence properties of the MOEAs. It proposes a simple and pragmatic MOEA model based on grids. The convergence of MOEAs is defined and the general conditions of convergence are provided. It also shows that the proposed (μ+1)-MOEA strongly converges to Pareto optimal set with probability one under suitable condition. Numerical results illustrate that this algorithm is feasible and effective.
出处
《计算机学报》
EI
CSCD
北大核心
2004年第10期1415-1421,共7页
Chinese Journal of Computers
基金
国家"八六三"高技术研究发展计划基金 (2 0 0 2AA1Z14 90 )
广东省自然科学基金博士启动项目基金 (0 43 0 0 15 7)资助
关键词
多目标
演化算法
收敛性
群体搜索
优化
evolutionary algorithms
multi-objective
optimization
convergence
