基于特征方程的蝙蝠算法分析及其改进策略唐建新(1.兰州大学信息科学与工程学院,甘肃兰州730000)摘要:在求解复杂非线性优化问题时,蝙蝠算法因其进化机制中引入了更多可调参数因子而比粒子群算法和遗传算法等具有更好的收敛性能。然而,在其迭代过程中,一旦群体中出现“超级”蝙蝠个体,算法极易出现“迟滞”问题。本文采用特征方程方法对基本蝙蝠算法的收敛性进行了分析,在一定假设条件下,讨论了算法参数灵敏性。基于负梯度理论,通过调整算法中蝙蝠个体的速度更新策略,使其沿群体当前最优解的负梯度方向飞行,引导个体飞向全局最优解。典型benchmark函数仿真实验结果表明,改进蝙蝠算法表现出较基本蝙蝠算法更好的全局寻优能力。关键词:最优化;蝙蝠算法;特征方程;收敛性分析;元启发式算法中图分类号:TP301文献标志码:AThetheoreticanalysisofbatalgorithmbasedoncharacteristicequationanditsimprovementstrategiesWANGXin1,TANGJian-xin2(1.DepartmentofInformationEngineering,GansuVocationalandTechnicalCollegeofCommunication,Lanzhou730070,China;2.SchoolofComputerandCommunication,LanzhouUnivofTech,Lanzhou730050,China)Abstract:Thebatalgorithm(BA)achievessatisfiedresultscomparingtoparticleswarmoptimization(PSO)andgeneticalgorithm(GA)insolvingcomplexnonlinearoptimizationproblemsfortherearemorecontrollableparametersincludedintoitsevolutionarymechanism.However,BAtendstobetrappedintoprematureeasilyonceasuperiorbatexploitedlocallyduringtheiterationprocesses.CharacteristicequationmethodwasemployedtoanalyzetheconvergenceoftheoriginalbatalgorithmandthesensibilityofBAparameterswasdiscussedbasedonpostulatedconditions.Animprovedbatalgorithmwaspresented,inwhichmodifiedvelocityupdatingstrategiesaccordingtonegativegradientdirectiontheoreticalanalysiswereintroduced,Typicalsimulatedexperimentalresultsindicatethevalidityandsuperiorityofthemodifiedmetaheuristicbatalgorithm.Keywords:Optimization;BatAlgorithm;CharacteristicEquation;ConvergenceAnalysis;MetaheuristicAlgorithm0引言最优化理论的目标之一是寻求最优化方法以求解复杂系统中的最优化问题。传统优化方法如单纯形法、整数规划等确定性算法因依赖待优化问题的结构信息而无法满足求解目前较为普适的非...