多阶段自适应差分进化算法及应用分析

《多阶段自适应差分进化算法及应用分析》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、东华人学管J里学博卜学位论文基于改进差分的聚类算法，用于提高数据聚类的准确性和质量。最后，对全文研究内容进行了总结，J一旨出研究工作中存在的不足，明确了下一步的研究方向。关键词:差分进化算法;阶段交叉;两段式;自适应;流水车间;二次规划;聚类AbstraetAbstractDifferentialevolution(DE)algorithmhasdemonstratedareliable，aecurate，robustandfastevolutionaryalgorithmandsueeessfulaPPliedinmanyreal一worldareas.H

2、owever，DEhasshownsomeweaknesses，esPeeiallyDEinsolvillgcomPlexfunctionswithhighdimensionshasshownsomeweaknesses，suchasloweonvergeneesPeed，PrematureeonvergeneeandlowPreeision，WithaneyetoimProvethePerformanceofDE，it15neeessarytodePenduPonaPProPriatelyehoosingeontrolParameters.Inviewof

3、thisProblemdePendeney，anovelself-adaptingdifferentialevolutionalgorithmwithPieeewiseoPerators15ProPosed.OPeratorsareresPeetivelyself-adaPtedbasedonCauehyorxlonnaldistributionrandomnumber，anddynamieallygeneratetheseeontrolParametersinthisalgorithm.TheevolutionProeess15dividedintotwo

4、stages，differenteonfigurationParametersareusedtogenerateamPlificationfactorsanderossoverProbabilitybyresPeetiveoPerators(ortoehoosesuitableerossoverProbability)ateaehstage.Atthesametime，toenhancetheeonvergentsPeed，anewmutationstrategy15izltrodueedtoguidesearehingdirection.Benelllna

5、rkProblemsareusedtoverifythisalgorithmandtheresultofsimulationindieatedthathasimProvedsignificantlybothineonvergeneesPeedandPreeisionofoPtimization.Inthisstudy，differentialevolutionaryeomPutinganditsaPPlieationarefocusedon.Somenovelself-adaPtivealgorithmsareProPosedbasedonmulti一sta

6、geandtraditionaldifferentialevolution.TorePairdeficieneies，suehasProblemdePendeney，lowPrecisioninsolvingeomPlexfunetionswithhighdimensionsandeasilytraPPedintoaloealoPtimum，seglllentalwaveandPieeewiseerossoveralgorithmsareProPosed.InordertoimProvethePraetiealityofthealgorithm，asolut

7、iontodealwiththeeonstrainedoPtimizationProblems15alsoProvided.TheProPosedalgorithmsfinallyapplyinthefieldssuehastransportationproblems，quadraticprograms，bloekingflowshoPschedulingandelusteringProblems.Moredetailsaregivenasfollow:(1)InsolvingasPeeifieProblemeruciallydePendsonaPProPr

8、iatelyehoosingeontrolParam