基于EDA和自適應(yīng)進(jìn)化策略的多目標(biāo)優(yōu)化方法研究.pdf

1、創(chuàng)新性聲明本人聲明所呈交的論文是我個(gè)人在導(dǎo)師指導(dǎo)下進(jìn)行的研究工作及取得的研究成果。盡我所知，除了文中特別加以標(biāo)注和致謝中所羅列的內(nèi)容以外，論文中不包含其他人已經(jīng)發(fā)表或撰寫過的研究成果；也不包含為獲得西安電子科技大學(xué)或其它教育機(jī)構(gòu)的學(xué)位或證書而使用過的材料。與我一同工作的同志對本研究所做的任何貢獻(xiàn)均已在論文中做了明確的說明并表示了謝意。申請學(xué)位論文與資料若有不實(shí)之處，本人承擔(dān)一切相關(guān)責(zé)任。本人簽名：關(guān)于論文使用授權(quán)的說明本人完全了解西安電

2、子科技大學(xué)有關(guān)保留和使用學(xué)位論文的規(guī)定，即：研究生在校攻讀學(xué)位期間論文工作的知識產(chǎn)權(quán)單位屬西安電子科技大學(xué)。本人保證畢業(yè)離校后，發(fā)表論文或使用論文工作成果時(shí)署名單位仍然為西安電子科技大學(xué)。學(xué)校有權(quán)保留送交論文的復(fù)印件，允許查閱和借閱論文；學(xué)?？梢怨颊撐牡娜炕虿糠謨?nèi)容，可以允許采用影印、縮印或其它復(fù)制手段保存論文(保密的論文在解密后遵守此規(guī)定)。本學(xué)位論文屬于保密在年解密后適用本授權(quán)書。本人簽名：蘭竺魚導(dǎo)師簽名：i燮：=：≯孓，≯；0

3、AbstractRecentlysomenewevolutionaryparadigmshavebeenintroducedintomulti—objectiveoptimizationproblem(EMO)communitysuchasartificialimmunesystem，estimationdistributionalgorithms，andMemeticalgorithmsThisarticlefocusedonabov

4、eseveralclassicalalgorithmanalysisandresearch，weproposedtwomuffobjectiveoptimizationalgorithmsbasedonEstimationofDistributionAlgorithm(EDA)andadaptiveevolutionstrategyARegularityModelBasedMultiobjectiveEstimationofDistri

5、butionAlgorithm(RM—MEDA)isoneofthedistributionalgorithmsItcanbeinducedthattheParetoset，inthedecisionspace，ofacontinuousmulti—objectiveoptimizationproblemshowsomecertainregularityBasedonthisregularitypropertyRM—MEDAbuilda

6、piecewiselinermodelandthensamplenewtrialsolutionsfromthemodelthusbuiltRMMEDAhasshowedtheobviousadvantageinsolvingonasetoftestinstanceswithvariablelinkagesHoweverRM—MEDAalsohassomedrawbacksFirstlydistributedestimationalgo

7、rithmonlygeneratesthenewsolutionsthroughsampling，butfailtousethelocationinformationofexistingoptimalsolutionsforlocalsearchThus，themethodthatadoptstheprobabilitymodelsamplingstrategytoproduceoffspringislesseffectivewhent

8、hepopulationdidn’tshowthecertainregularityinthedistributionofpopulmionSecondlythereasonablenumberofclustersofthepopulationisrelatedtotheParetooptimalsolutionsetofthespecifictestquestions，butinRMMEDA，thenumberofclusternee

9、dtoartificiallysetthenumberofclustersKTosolvetheaboveproblems，themainresearchworkofthispaperincludesthefollowing：(1)BasedonEstimationofDistributionAlgorithm(EDA)andImmuneClonalSelectionAlgorithm(ICSA)，thispaperproposedaM

10、emeticAlgorithmwithAdaptiveOperatorSelectionformutiobjectiveoptimization(MAOS)InMAOS，twotypesofreproducingstrategiesarecombined：oneistherecombinationandmutationbasedimmuneclonalselectionoperatortheotherisEDAbasedmodeling

11、andsamplingoperatorInMAOS，theadaptiveMemeticoperatordefinitetwoindicators：capacityandefficiencybyusingthehistoricalinformationofeliteindividualstoadaptivelyadjustbothoperatorsusedindifferentstagesofevolution，thusMAOSCanb

12、alancetheexploimtionandexplorationofthealgoritbanExperimentalstudieshaveshownthattheadaptiveMemeticoperatorwasefficientAndMAOSperformsbetterthanRM—MEDA，NSGAIIonasetoftestinstanceswithlinearandnonlinearvariablelinkagesand