混合蛙跳算法的收敛性分析及其改进  

Convergence analysis of shuffled frog leaping algorithm and its modified algorithm

作　　者：肖莹莹[1] 柴旭东[2] 李伯虎[1,2] 王秋生[1] 

机构地区：[1]北京航空航天大学自动化与电气工程学院,北京100083 [2]北京仿真中心,北京100854

出　　处：《华中科技大学学报（自然科学版）》2012年第7期15-18,28,共5页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家高技术研究发展计划资助项目(2011AA040502)

摘　　要：在分析混合蛙跳算法(SFL)收敛性的基础上,针对早熟收敛和收敛速度慢的问题,提出一种改进算法(MSFL).MSFL利用变公比数列分析更新轨迹的收敛性,并引入离散度和适应度方差作为指标,自适应地调节数列公比取值范围,以平衡收敛精度和收敛速度.以6个Benchmark函数分2组实验,测试MSFL的性能.结果表明:提出的MSFL算法具有较强的全局搜索和局部搜索能力,且具有收敛速度快、收敛精度高的优点.Based on the analysis of the conditions of shuffled frog leaping（SFL）algorithm,a modified SFL（MSFL）was proposed to deal with the premature and slow convergence.In MSFL,convergence was analyzed based on the theory of geometrical sequence,and two premature judgments were defined,including the dispersion and the fitness variance,to adaptively adjust the coefficients,for balancing convergence accuracy and rate.Six benchmark functions were proposed to test the performances of MSFL,and experiment simulations show that MSFL has higher precision and good stability.

关 键 词：混合蛙跳算法 收敛性分析 更新策略 等比数列 多目标优化问题 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]