摘要
分布性保持是多目标进化算法主要目标之一.然而通常维护方法的性能与运行时间存在矛盾.提出一种基于最小生成树的分布性维护方法.利用最小生成树中的度数和边长对个体密度进行估计,使低度数的边界个体和长边长的低密度个体得到了保留.另外,一次性选择个体进入下代种群,避免了每移出一个个体就需要对个体密度进行调整的操作.通过5个测试问题和4个方面的测试标准,与3个著名的算法进行比较实验,结果表明该方法在以较快速度对种群进行维护的同时,拥有良好的分布性.
Maintaining the diversity of solutions is a crucial part in multi-objective optimization. However, there has to be a trade-off between the diversity and the execution time. A method for maintaining the diversity using a minimum spanning tree is proposed. By estimating the individual density based on the degree and edge of the minimum spanning tree, we preserve the low-degree boundary individuals and the longer-edge-low-density individuals. Moreover, by this one- time selection, the adjustment of individual density after removing each individual can be avoided. Through the extensive comparison study with three other classical methods on four performance metrics in five test problems, it is observed that the proposed method has a good performance in diversity and execution time.
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
2009年第8期843-849,共7页
Control Theory & Applications
基金
国家自然科学基金资助项目(60773047)
留学回国人员科研启动基金资助项目(教外司留[2005]546号)
湖南省自然科学基金资助项目(05JJ30125)
湖南省教育厅重点科研资助项目(06A074)
关键词
多目标优化
进化算法
分布性维护
最小生成树
multi-objective
evolutionary algorithms
diversity maintenance
minimum spanning tree