Reevaluating Genetic Algorithm Performance under coordinate Rotation of Benchmark Function

《Reevaluating Genetic Algorithm Performance under coordinate Rotation of Benchmark Function》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、ReevaluatingGeneticAlgorithmPerformanceundercoordinateRotationofBenchmarkFunctionsAsurveyofsometheoreticalandpracticalaspectsofgeneticalgorithmsRalfSalomonAILab,ComputerScienceDepartment,UniversityofZurichWinterthurerstrasse190,8057Zurich,SwitzerlandFAX:+41-1-3630035;Emai

2、l:salomon@i.unizh.chAbstractThisworkanalyzessomeconceptsofgeneticalgorithmsandexplainswhytheymaybeappliedwithsuccesstosomeproblemsinfunctionoptimization.Inadditiontootherperformanceproperties,ithasbeenshownthatgeneticalgo-rithmsareabletoovercomelocalminimainhighlymultimo

3、dalfunctions(e.g.,Rastrigin,Schwefel).Theperformanceofgeneticalgorithmsissupportedbyanextensivetheory,whichisbasedontheassumptionofadditivegeneeects.Butthecurrentworkshowsthattheassumptionofadditivegeneeectsisnotweak,andthatthedependenceonspecicparametersettingsismuchs

4、trongerthanoftenbelieved.Furthermore,theassumptionsregardingthetnessfunctionaresorestrictingthatslightmodicationsofthestandardtestfunctionscauseafailureoftheoptimizationprocedureeventhoughthefunction'sstructureispreserved.Thecurrentexperimentsfocusonafewwidely-usedscala

5、bletestfunctions.theresultsindicatethatastandardgeneticalgorithmcanndtheglobaloptimumofRastrigin-likefunctionsnotduetotheeciencyofthere-combinationoperatorandtheregulardistributionoflocalminima,butbecausetheoptimizationisdecomposableintonindependentone-dimensionalsubpro

6、b-lems.Asalogicalconsequenceoftheanalysis,itisshownhowthefunction'spropertiescanbeexploitedinordertoconstructamoreecientalgorithm.KeywordsGENETICALGORITHMS{FUNCTIONOPTIMIZATION{PERFORMANCECOORDINATEROTATION11IntroductionAgeneticalgorithm(GA)isaheuristicsearchprocedurebas

7、edonnaturalselectionandgenetics,whichmaintainsapopulationoftrailsolutions.Geneticalgorithms(GAs)aresuitableforaddressingmanyfunctionoptimizationproblems.Recentde-velopmentssuchastheparallelgeneticalgorithm[10]andthebreedergeneticalgo-rithm[12,13]havebeenshowntobesuccessfu

8、linoptimizingmultimodalfunctions.IthasbeenreportedthatsuchGAsovercomemillionsoflocaloptimaandcan