一种基于ε-支配的多目标自组织迁移算法  

A new ε-dominance based multi-objective self-organizing migrating algorithm

作　　者：林志毅[1] 王玲玲[2] 

机构地区：[1]广东工业大学计算机学院,广州510006 [2]武汉大学软件工程国家重点实验室,武汉430072

出　　处：《控制与决策》2014年第12期2177-2182,共6页Control and Decision

基　　金：广东高校优秀青年创新人才培养计划项目(2012LYM 0054);广州市科技计划项目(2012Y2-00046;2013Y2-00043)

摘　　要：提出一种基于ε-支配关系的多目标自组织迁移算法.首先对基于混合迁移行为的自组织迁移算法进行改进,将全面学习的思想引入个体的迁移过程中,使得个体的每个分量都可以向其他个体学习,从而进行充分的信息交换;通过引入学习因子扩展个体的迁移方向,使得步长可灵活变化,进而利用改进算法快速搜索多目标函数的Pareto最优解,并采用ε-支配关系以保持种群分布性;最后通过实验分析表明了所提出方法的有效性.A ε-dominance based multi-objective self-organizing migrating algorithm（EMSOMA） is proposed. Firstly, a hybrid migrating behavior based self-organizing migrating algorithm（HBSOMA） is improved to quickly search the Pareto optimal solutions of the multi-objective optimization problems. The improved HBSOMA introduces the idea of comprehensive learning into the migration process, in which each component of individual can learn from different individuals for exchanging information adequately. Moreover, a learning factor is employed in the migration process to extend the migrating direction and change the step size flexibly. And based on the ε-dominance relationship, EMSOMA can maintain the diversity of the population. Finally, the experiment analysis shows the effectiveness of the proposed method.

关 键 词：多目标优化 ε-支配 自组织迁移算法 ε-支配关系的多目标自组织迁移算法 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]