一种新的群搜索优化实现算法  

A Novel Realization Algorithm of Group Search Optimizer

作　　者：罗磊[1] 谢静[1] 周晖[1] 梁天[1] 冯绍杰[1] 庆栋良[1] 

机构地区：[1]南通大学电子信息学院,江苏南通226019

出　　处：《南通大学学报（自然科学版）》2012年第2期1-8,共8页Journal of Nantong University(Natural Science Edition)　

基　　金：国家自然科学基金资助项目(60901041);江苏省2011年度普通高校研究生科研创新计划项目(CXLX11_0647);江苏省高校社会科学研究项目(2011SJB630048);南通大学研究生科技创新计划项目(YKC11056;YKC10023)

摘　　要：针对群搜索优化(GSO)算法存在的不足,提出一种新的GSO实现算法(NRGSO).采用5个300维和7个30维的测试函数对NRGSO算法进行数值实验,并将其与GSO算法、微粒群优化(PSO)算法、遗传算法(GA)、进化规划(EP)、进化策略(ES)进行比较.结果表明,NRGSO算法的性能优于GSO算法;而在解决高维和多模态函数的优化问题方面,其性能优于PSO、GA、EP和ES等算法.NRGSO算法改进了群搜索优化原实现方法的不足,提高了算法的搜索性能,不仅在高维函数的优化中表现卓越,还能有效地避免陷入局部次优,并且在实际的优化问题中应用方便.A novel realization algorithm of group search optimizer （NRGSO） is proposed, aiming at overcoming the deficiency of GSO. And it is easier to be applied in practical problems. Five test functions of 300 dimensions and seven test functions of 30 dimensions are used to conduct the numerical experiments and the results of the novel algo-rithm are compared with those of GSO, particle swarm optimization （PSO）, genetic algorithm （GA）, evolutionary programming （EP） and evolutionary strategy （ES）. The algorithm proposed in this paper is better than GSO and its performance in solving the problems of high dimensions and multimodal functions is better than PSO, GA, EP and ES. NRGSO improves the original algorithm. It enhances its search ability and achieves better results. This novel algorithm performs excellently in functions of high dimensions, can effectively avoid being trapped in the local minima and is applicable in practical optimizer.

关 键 词：群搜索优化 函数优化 多模态函数 高维函数 算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]