基于合作协同策略的演化算法分析

《基于合作协同策略的演化算法分析》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、ABSTRACTAbstractOptimalproblemisallimportantissueandwildlyexistedinourdailylifeandwork．Asallefficientheuristicoptimizationalgorithm，evolutionaryalgorithmhasbeengetgreatprogressanddevelopedaseriesofnewalgorithmsandtechniques，suchasGP，DE，PSO，ACOandSOon．Th

2、oseclassicalgorithmscallreachbetterresultinlow—dimensionproblemsandtheycanfindthelastoptimalsolutioncorrectlyandquicklyiftherearelittervariables．Butwiththevariablesnumbergrowth，theproblemgetcomplex，manyalgorithmswhichhaveexcellentperformanceinlow．dimens

3、ionproblemslosttheirperformance．Inactualproject，manyoptimalproblemshavelargescale,andmayalsoexistingvariable-correlationproblem．Forexample，density,consumptionandheatdissipationproblemsexistinthedesigningofIntegratedCircuit(IC)，whichalsoneedpowerfuloptim

4、izationalgorithms．Recentyears，manyresearchesinhighdimensionalnumericaloptimizationproblemsmainlyconcentratedinthevariablecorrelation，solutionspacereduce，evolutionstatejudgment，algorithmselectingandusingofadaptivemethods．Thisthesismainlyfocusesonthedesig

5、nforlargescaleglobaloptimizationproblems．Inthispaper，basedonparticleswarmoptimization(PSO)algorithmandcooperativeCO‘evolution，westudiedsometechniquesinlargescaleglobaloptimization．Themainworkandtheinnovationaresummarizedasfollows：1．Studiedcooperativeco-

6、evolutionbasedparticleswarmoptimization(CCPSO)algorithm，analyzedtheperformanceofCCPSOinlargescaleglobaloptimization．．2．WhenusingCCPSOalgorithm，wetrackedtheparticlevelocityandfoundaphenomenonthatthevelocitiesofalltheparticlesofadimensionbecomedivergentwh

7、entherealoptimalsolutionofsuchdimensioniSneartheboundofthesearchspace．ThedivergenceOfvelocitywilldecreasetheconvergencespeedofalgorithm．Basedonthisobservation，proposedvelocitydivergencedetectionandre—initializationtechniquedealingthedivergencevelocityan

8、dachievedgoodresults．3．Studiedthetechniqueofcombinationmultievolutionaryalgorithms．Basedon“nofreelaunch”，usingcompetitivelearningmethod，trackedtheperformanceofeachoffspringalgorithm．Goodperformancealgorithmshavemoreprobabilitytop