基于种群自适应策略的差分演化算法及其应用研究

《基于种群自适应策略的差分演化算法及其应用研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、AOsU'aetegyisproposedtoplaceseveralnewindividualsinappropriateareastodiscovernewpos—siblesolutions．Meanwhile，allinferior-basedpopulation—cutstrategyisalsopresentedtoremoveseveralpoorparticlesaccordingtoitsrankingmethodandtoreserveaplaceforabetterreproduction．Moreover,bot

2、lldynamicpopulationstrategiesarecontrolledbyastatusmonitor,whichisusedtokeeptrackoftheprogressofindividualsandimprovethesensitivityoftheproposedAPTS．neexperimentalstudieswerecarriedouton25globalnumericaloptimizationproblemsusedintheCEC2005specialsessiononreal—parameterop

3、timization．AnoverallperformancecomparisonbetweentheJADE-APTSvariantandotherfiveState—of-the—ArtDEswasalsocarriedout．TheexperimentalresultsillustratedthatJADE-APTSachievesacompetitiveperformancein30dimensionalproblemsandexhibitsthebestperformancein100dimensionalproblems．I

4、naddition，theANOVAresultsverifythatAPTScanacceleratetheconvergenceandenhanceaccuracy．(2)Enhanceddifferentialevolutionwithentropy-basedpopulationadapta-tionandmarkovchainmodelAnenhancedadaptivepopulation-handlingtechnique(CP)isproposedforDEalgo。rithmtosolvevarioustypesofo

5、ptimizationproblems．InCPDE，weadvocateastochasticstrategy—hoppingframeworkinwhichtheprobabilityofselectingdifferentsub—optimizerstoimprovetheonlinesolution-searchingstatusiscompletelyfollowedbyaMarkovchain．Onesub·optimizer,calledpopulationincreasingstrategy,addsnewindivid

6、ualsintothepopulationtosharetheirup·-to··dateinformationwhenparticlesareclusteredtogetherinaregionandtrappedintothelocalbasin；theothersub—optimizer,namelypopulationde‘creasingstrategy，removesredundantparticlewithitsentropyandrankingmetricstosavecomputationalload．Extensiv

7、eexperimentshavebeencarriedouttocompareitwithfivestate—of-the．artDEvariantsandthreeotherEAson25commonlyusedCEC2005con—testinstances．Inaddition，ascalabilitystudywasimplementedtOshowtheeffectofproblemdimension．Intheend，mntimecomplexityanalysisandconvergenceratecorn—parison

8、arealsovalidatedthatCPframeworkdoesnotimposeanyseriousburdenonthetimecomplexityoftheexistingDEvariants．