资源描述：
《经济学--Improved Genetic Algorithm and Its Performance Analysis》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、TransactionsofTianjinUniversityVol.9No.2Jun.2003GeneticAlgorithmandItsPerformanceAnalysisLUO。。，，，，:，、。，。，，GUOJi-chang。，，，，，TENGTJaI88-004M)SchoolofElectronicInformationEngineering,TianjinUniversity,Tianjin300072,China)Abstract;Althoughgeneticalgorithmhasbecomeveryfamou
2、swithitsglobalsearching,parallelcomputing,betterrobustness,andnotneedingdifferentialinformationduringevolution.However,italsohassomedemer-its,suchasslowconvergencespeed.Inthispaper,basedonseveralgeneraltheorems,animprovedgeneticalgorithmusingvariantchromosomelengthandp
3、robabilityofcrossoverandmutationisproposed,anditsmainideaisasfollows;atthebeginningofevolution,oursolutionwithshorterlengthchromosomeandhigherprobs-bilityofcrossoverandmutation;andatthevicinityofglobaloptimum,withlongerlengthchromosomeandto,,erprobabilityo#crossoverand
4、mutation.Finally,testingwithsomecriticalfunctionsshowsthatoursolutioncanimprovetheconvergencespeedofgeneticalgorithmsignificantly,itscomprehensiveperformanceisbetterthanthatofthegeneticalgorithmwhichonlyreservesthebestindividual.variantchromosomelength;variantprobabili
5、ty;geneticalgorithm;on-lineandoff-lineperformArticleIDa1n00t6e4982(2003)02-0140-04Geneticalgorithmisanadaptivesearchingtechniqueeraltheorems(seeappendix)asfollows;Letusassumebasedonaselectionandreproductionmechanismfoundinthereisjustonevariable(multi-variablecanbedivid
6、edin-thenaturalevolutionprocess,anditwaspioneeredbyHol-tomanysections,onesection阮onevariable)、二厂a,b]，landinthe1970s.IthasbecomeveryfamouswithitsxER,andchromosomelengthwithbinaryencodingis1.globalsearching,parallelcomputing,betterrobustness,Theorem1Minimalresolution可chr
7、omosomeisandnotneedingdifferentialinformationduringevolution.b一aHowever,italsohassomedemerits,suchaspoorlocal“-2花1searching,prematureconverging,aswellasslowconver-Theorem2Weightvalueoftheithbitofchromo-gencespeed.Inrecentyears,theseproblemshavebeensomeisstudied"_'一二_2b
8、`--a12`(i二112,...,1)Inthispaper,animprovedgeneticalgorithmwithva-riantchromosomelengthandvariantprobabilityispro-Theo
