资源描述：
《a tutorial on evolutionnary multiobjective optimization》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、ATutorialonEvolutionaryMultiobjectiveOptimizationEckartZitzler,MarcoLaumanns,andStefanBleulerSwissFederalInstituteofTechnology(ETH)Zurich,ComputerEngineeringandNetworksLaboratory(TIK),Gloriastrasse35,CH-8092Zurich,Switzerland{zitzler,laumanns,bleuler}@tik.ee.ethz.chAbstract.Multiple,oftenconflictingo
2、bjectivesarisenaturallyinmostreal-worldoptimizationscenarios.Asevolutionaryalgorithmspossessseveralcharacteristicsthataredesirableforthistypeofproblem,thisclassofsearchstrategieshasbeenusedformultiobjectiveoptimizationformorethanadecade.Meanwhileevolutionarymultiobjectiveoptimiza-tionhasbecomeestabl
3、ishedasaseparatesubdisciplinecombiningthefieldsofevolutionarycomputationandclassicalmultiplecriteriadecisionmaking.Thispapergivesanoverviewofevolutionarymultiobjectiveoptimiza-tionwiththefocusonmethodsandtheory.Ontheonehand,basicprin-ciplesofmultiobjectiveoptimizationandevolutionaryalgorithmsareprese
4、nted,andvariousalgorithmicconceptssuchasfitnessassignment,diversitypreservation,andelitismarediscussed.Ontheotherhand,thetutorialincludessomerecenttheoreticalresultsontheperformanceofmultiobjectiveevolutionaryalgorithmsandaddressesthequestionofhowtosimplifytheexchangeofmethodsandapplicationsbymeansof
5、astandardizedinterface.1IntroductionThetermevolutionaryalgorithm(EA)standsforaclassofstochasticoptimiza-tionmethodsthatsimulatetheprocessofnaturalevolution.TheoriginsofEAscanbetracedbacktothelate1950s,andsincethe1970sseveralevolution-arymethodologieshavebeenproposed,mainlygeneticalgorithms,evolution
6、aryprogramming,andevolutionstrategies[1].Alloftheseapproachesoperateonasetofcandidatesolutions.Usingstrongsimplifications,thissetissubsequentlymodifiedbythetwobasicprinciples:selectionandvariation.Whileselectionmimicsthecompetitionforreproductionandresourcesamonglivingbeings,theotherprinciple,variatio
7、n,imitatesthenaturalcapabilityofcreating”new”livingbeingsbymeansofrecombinationandmutation.Althoughtheunderlyingmechanismsaresimple,thesealgorithmshaveproventhemselvesasageneral,robustandpowerfulsearc
