Gary-G.-Yen教授的報(bào)告:多目標(biāo)粒子群算法_第1頁(yè)
Gary-G.-Yen教授的報(bào)告:多目標(biāo)粒子群算法_第2頁(yè)
Gary-G.-Yen教授的報(bào)告:多目標(biāo)粒子群算法_第3頁(yè)
Gary-G.-Yen教授的報(bào)告:多目標(biāo)粒子群算法_第4頁(yè)
Gary-G.-Yen教授的報(bào)告:多目標(biāo)粒子群算法_第5頁(yè)
已閱讀5頁(yè),還剩36頁(yè)未讀 繼續(xù)免費(fèi)閱讀

下載本文檔

版權(quán)說(shuō)明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請(qǐng)進(jìn)行舉報(bào)或認(rèn)領(lǐng)

文檔簡(jiǎn)介

MANY-OBJECTIVEPARTICLESWARMOPTIMIZATIONBASEDONPARALLELCELLCOORDINATESYSTEMGaryG.Yen,Ph.D.,FIEEE,FIETProfessor,OklahomaStateUniversityPastPresident,IEEEComputationalIntelligenceSocietyFoundingEditor-in-Chief,IEEEComputationalIntelligenceMagazine

PaperSubmissionDeadline:December20,2013July24-29,2016MultiobjectiveOptimization

OptimizationproblemsinvolvemorethanoneobjectivefunctionVerycommon,yetdifficultproblemsinthefieldofscience,engineering,andbusinessmanagementNonconflictingobjectives:achieveasingleoptimalsolution satisfiesallobjectivessimultaneouslySOPsCompetingobjectives:cannotbeoptimizedsimultaneouslyMOP–searchforasetof“acceptable”–maybeonlysuboptimalforoneobjective–solutionsisourgoalInoperationresearch/managementterms–multiplecriteriondecisionmaking(MCDM)WhyMOP?BuyinganAutomobileObjective=reducecost,whilemaximizecomfortWhichsolution(1,A,B,C,2)isbest???NosolutionfromthissetmakesbothobjectiveslookbetterthananyothersolutionfromthesetNosingleoptimalsolutionTradeoffbetweenconflictingobjectives-costandcomfortMathematicalDefinitionMathematicalmodeltoformulatetheoptimizationproblem

DesignVariables:decisionandobjectivevectorConstraints:equalityandinequalityGreater-than-equal-toinequalityconstraintcanbeconvertedtoless-than-equal-toconstraintbymultiplying-1ObjectiveFunction:maximizationcanbeconvertedtominimizationduetothedualityprincipleObjectivevectorsDecisionvectorsEqualityconstraintsInequalityconstraintsVariableboundsEnvironmentstatesParetoOptimalityFormalDefinition:theminimizationofthencomponents

ofavectorfunctionfofavectorvariablexinauniverse,whereThenadecisionvectorissaidtobePareto-optimalifand onlyifthereisnoforwhich dominates,thatis,thereisnosuchthatNondominatedset(Paretofront)f1f2objectivespaceABCParetoOptimalSet–thesetofallPareto-optimaldecisionvectors,whichyieldsasetofnondominatedsolutionsNon-dominatedSet–correspondingobjectivevectorset-ParetoFrontx2x1ParetooptimalsetABCdecisionspaceDZDTTestSuiteAnunorthodox,stochastic,andpopulationbasedparallelsearchingheuristicsmaybemoresuitableforMOPsClassificationofEA’s–GeneticAlgorithm;GeneticProgramming;EvolutionaryProgramming;EvolutionaryStrategy;AntColony;ArtificialImmuneSystem;ParticleSwarmOptimization;DifferentialEvolution;MemeticAlgorithmWhyPopulation-BasedHeuristics?abilityofhandlingcomplexproblemswithdiscontinuities,multimodality,disjointfeasiblespacesanddynamismResearchIssuesforMOPsModifyingthefitnessassignmentEnhancingtheconvergencePreservingthediversityManagingthepopulationConstraintsanduncertaintyhandlingProgressionsofdevelopmentinEMOcommunity-

fromevolutionary&nature-inspiredcomputationalmetaphors,

tosolvingsingle-objective

optimizationproblems,

tosolvingconstrained

optimizationproblems,

tosolvingdynamic

optimizationproblems,

tosolvingmulti-objectiveoptimizationproblems,

andtosolvingnowMany-ObjectiveOptimizationProblems.Multi-ObjectiveOptimizationProblems(MOPs)withalargenumberofobjectives(ingeneraloverfive)arereferredtoasMaOPs.

ProgressioninEMODevelopmentsWhenencounterproblemswithmanyobjectives(morethanfive),nearlyallalgorithmsperformspoorlybecauseoflossofselectionpressureinfitnessevaluationsolelybaseduponParetodomination.Withtheincreasingnumberofobjectives,thereareafewchallengestobeaddressed:IneffectivedefinitioninthePareto-dominancedeterioratestheconvergenceabilityofMOEAsAnexponentiallylargenumberofsolutionsarerequiredtoapproximatethewholeParetofrontInbalanceofcomputationalcomplexityandqualityofthesolutionfoundVisualizationofalarge-dimensionalfrontisreallydifficultMetricstoquantifytheperformanceofthedesignsResearchIssuesforMaOPsObjectiveReductionNon-Pareto-BasedTechniquesIncorporationofPreferenceInformationGradientInformationModificationsofMOEAsforMaOPsParetoDominationRevisionsDominanceAreaControl,?-Dominance,k-Optimality,GridDominance,Fuzzy-basedParetoDominanceFD-NSGA-II,FD-SPEA2(He&Yen,TEVC,2013)DecompositionMethodsMOEA/D(Zhang&Li,TEVC,2007);NSGA-III(Deb&Jain,TEVC,2013)GridBasedApproachesTDEA(Pierroetal.,TEVC,2007);e-MOEA(Debetal.,EvolComput,2005);GrEA(Yangetal.,TEVC,2013)IndicatorBasedMethodsSMS-EMOA(Beumeetal.,EJOperResearch;2007)HypE(Bader&Zitzler,EvolComput.,2011)State-of-the-ArtMaOEAsInPSOside…Meta-heuristicallyinspiredbythesocialbehaviorofbirdflockingorfishschooling,therelativesimplicityandthesuccessasasingle-objectiveoptimizerhavemotivatedresearcherstoextendPSOfromSOPstoMOPs.However,apartfromthecommonissueinMOEAstomaintainanarchive,therearetwoparticularissuesinMOPSO:ManagingconvergenceanddiversityfastconvergenceofPSOincursarapidlossofdiversityduringtheevolutionaryprocessSelectingglobalbest(gBest)andpersonalbest(pBest)thereisnoabsolutebestsolutionbutratherasetofnon-dominatedsolutions.ManymechanismswereproposedintheexistingMOPSOsintermofleaderselection,archivemaintenance,andperturbationtotackletheseissues.However,fewMOPSOsaredesignedtodynamicallyadjustthebalanceinexplorationandexploitationaccordingtothefeedbackinformationthroughinteractingtheevolutionaryenvironment.ThechallengesinMOPSOformanagingtheconvergenceanddiversity:updatingarchiveselectinggBestandpBestSelf-adaptingflightparametersperturbingstagnationMotivationsAmechanism(differentfromgrid-basedapproaches)for:assessingdiversitytoselectglobalbestforaparticleandupdatearchiveevaluatingtheevolutionaryenvironmenttodynamicallyadjusttheevolutionarystrategiesParallelcoordinates

isa

popularwayofvisualizing

high-dimensionalgeometry

andanalyzingmultivariate

data.Transformamulti-objective

spaceintoa2-Dgridto

evaluatethedistribution

ofanapproximateParetofrontParallelCellCoordinateSystemMaptheindividualsinglobalarchivefromCartesianCoordinateSystemintoParallelCellCoordinateSystem(PCCS)KbyMcellsKnon-dominatedindividualsinM-objectivespaceestimatingdensitytoupdatearchiveandselectdiversitygBest(d_gBest)withminimaldensityThedistancebetweentwovectors,namedParallelCellDistance,ismeasuredbythesumofnumbersofcellsawayfromeachotherinallobjectives.ThedensityofPi,inthehyperspaceformedbythearchivecanbemeasuredbytheParallelCellDistancebetweenPiandallothermembers,Pj(j=1,2,…,K,j≠i),inthearchive.rankingnon-dominatedsolutionsinarchiveforselectingconvergencegBest(c_gBest),withminimalpotential.c_gBestThepotentialquantifiesanon-dominatedsolutionamongitscompetitorsinthearchivebycombiningtheorderrelationalongtheoptimizationdirectionandthedegreeintheunitofcellinPCCS.DetectingtheEvolutionaryEnvironmentbyEntropyAbruptchangesindicateaconvergencestatusbecauseanewsolutiondominatessomeoldsolutionsinarchiveandthepopulationmakesaprogressorbreaksthroughalocalParetofront.Mildchangesindicateadiversitystatusbecauseanewsolutionwithbetterdensityreplacesanoldsolutioninarchive.Nochangeindicatesastagnationstatus.CurvesofEntropyandΔEntropydetectedfromZDT4withmanylocalParetofronts.ProposedpccsAMOPSOUpdatingarchiveComplexity:O(ML2)M:thenumberofobjectivesL:thenumberofMembersinarchiveSelectinggBestLeaderGroupMc_gBests&Md_gBeststhetypeofcandidatesbetweenc-gBestandd-gBestisdecidedinprobabilityacandidateforaparticleisrandomlydrawnfromthechosentypeofgBestAcandidateisrandomlydrawnfromthechosentypeThethresholdisthemaximalprobablevariationofentropy.SelectingpBestfrompArchivepArchivewithaquarterboundedsizeofgArchivetodecreasethecostofpArchivemaintenancepBestisselectedaccordingtotheminimalhyperbox:Self-AdaptivePSOflight-Anexampleofself-adaptiveparametersobtainedbypccsAMOPSOforZDT4withmanylocalParetofrontsPerturbingaparticletoacceleratetheconvergenceorescapefromlocalParetofrontsElitismLearningStrategy[fromZ.H.Zhan,J.Zhang,Y.Li,andH.S.Chung,“Adaptiveparticleswarmoptimization,”IEEETrans.Syst.Man,Cybern.B,Cybern.,vol.39,no.6,pp.1362-1381,Dec.2009.]RandomlyperturbadimensionofgBestPerturbationrangeisdampedbyaGaussianfunction,G(0,lr2)learningrate,lr,islinearlydecreasedfrom1.0downto0.1.TheintegratedalgorithmofpccsAMOPSOpccsAMOPSOforDTLZ2(3)ExperimentPeeralgorithmssigmaMOSPO:SigmavaluemethodbyMostaghimandTeich,2003agMOPSO:adaptivegridbyPadhye,2009cdMOPSO:crowdingdistancebyCoello,PulidoandLechuga,2004clusterMOPSO:clusteringpopulationbyMostaghimandTeich,2003pdMOPSO:ROUND+RAND+PROBbyAlvarez-Ben’itez.EversonandFieldsend,2005TestinstancesZTDseries(2-objective):fiveinstancesDTLZseries(3-objective):seveninstancesMetric:IGD&HyperVolume(HV)referencepointat11ineachobjectiveforHVHypervolume1-pccsAMOPSO2-sigmaMOPSO3-agMOPSO4-cdMOPSO5-clusterPOPSO6-pdMOPSO1-pccsAMOPSO2-sigmaMOPSO3-agMOPSO4-cdMOPSO5-clusterPOPSO6-pdMOPSOScoresofMeriton12testinstancesforHVProposedMOPSOsigmaMOPSOagMOPSOcdMOPSOclusterMOPSOpdMOPSONSGA-IISPEA2MOE

溫馨提示

  • 1. 本站所有資源如無(wú)特殊說(shuō)明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請(qǐng)下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請(qǐng)聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁(yè)內(nèi)容里面會(huì)有圖紙預(yù)覽,若沒(méi)有圖紙預(yù)覽就沒(méi)有圖紙。
  • 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
  • 5. 人人文庫(kù)網(wǎng)僅提供信息存儲(chǔ)空間,僅對(duì)用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對(duì)用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對(duì)任何下載內(nèi)容負(fù)責(zé)。
  • 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請(qǐng)與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶因使用這些下載資源對(duì)自己和他人造成任何形式的傷害或損失。

最新文檔

評(píng)論

0/150

提交評(píng)論