基于精英协同的混洗差分进化算法及其应用  

Shuffled Differential Evolution Algorithm Based on Elite Synergy and Its Application

作　　者：张大斌[1,2] 杨添柔[1] 潘玉辰[1] 周茜[1] 张文生[2] 

机构地区：[1]华中师范大学信息管理学院,湖北武汉430079 [2]中国科学院自动化研究所,北京100190

出　　处：《运筹与管理》2013年第5期17-23,共7页Operations Research and Management Science

基　　金：国家自然科学基金资助项目(70971052);中国博士后基金资助项目(2012M510607)

摘　　要：提出了基于精英协同的混洗差分进化算法(Shuffled Differential Evolution,SDE)。该算法引入反向学习的初始化机制,并对设置的普通群和虚拟精英群采用不同的差分策略,进而将精英个体作为信息通道实现种群间的信息交流;同时,借助定期混洗机制实现种群间的文化交流,从而达到协同进化的目的;此外,对长期停滞的个体进行跳变操作,以充分挖掘种群的搜索潜能,增强搜索的有效性。通过函数仿真,并与PSO及其它差分进化算法比较,结果表明该算法具有较好的寻优能力。This paper presents a novel Shuffled Differential Evolution algorithm （SDE）based on elite synergy. The algorithm introduces the initialization mechanism of opposition-based learning, employs different differential strategies fof several ordinary groups and a virtual elite group so as to take the elite individuals as the information channel for achieving information exchange among different groups. Meanwhile, it realizes the inter-cultural exchange among different groups by using a regularly shuffled mechanism which regroups the small groups via hash function, so as to achieve the population co-evolution. In addition, hopping operation on the individuals which are in the long-term stagnation can fully tap the potential of population search and enhance the effectiveness of the search. By the benchmark function experiments, the SDE performs better optimization capability in comparison with the Particle Swarm Optimization and other Differential Evolution algorithms.

关 键 词：最优化理论 差分进化 反向学习机制 协同机制 混洗思想 多种群 连续域问题 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]