evolutionary computation in constraint satisfaction and machine learning

《evolutionary computation in constraint satisfaction and machine learning》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、EvolutionaryComputationinConstraintSatisfactionandMachineLearningJanoI.vanHemertLeidenInstituteofAdvancedComputerScience,LeidenUniversityjvhemert@cs.leidenuniv.nl1IntroductionAtfirstsightthetwoproblemareasconstraintsatisfactionandmachinelearningdonotappeartobesimilar.Thefi

2、rsthasacleardefinitionofitsproblemdomainandacrispdefinitionofsolutions.Thesecondisamuchbroaderdefinedproblemdomain,whichleadstomanyobjectivestobesolved.Manyresearchareashavefocusedthereattentiononconstraintsatisfaction,operatingresearch,antcolonies,evolutionarycomputationan

3、d,mostnotable,constraintprogramming.Althoughtheproblemsthatarebeingstudiedsharethesamegoal,whichistosatisfyasetofconstraints,theirprecisedefinitionvaries.Amongtheseproblemswefindnumerouswellknownonessuchas,k-graphcolouring,3-satandn-queens.Foranyoftheseproblemswecantransfo

4、rmthemtoabinaryconstraintsatisfactionproblemwithoutlossofgenerality.Thisholdsforanyfiniteconstraintsatisfactionproblem,anditmeansthatwecansolveaproblembysolvingacorrespondingbinaryrepresentation[14],whichmeansthateveryconstraintonlyrestrictstheassignmentofvaluestotwovaria

5、bles.Inpractice,however,weneedtobecarefulasdifferentformsoftransformationarepossible,eachwithitsownimpactontheperformanceoftheappliedsolvingalgorithm.Incontrasttothefieldofconstraintsatisfactionthefieldofmachinelearn-ingtriestocopewithmanydifferentaspects.Therewefindproblemss

6、uchasclustering,classification,associationrulesandsymbolicregression.Oftentheseareliftedoutoneatatimeresultinginspecifictechniquestosolvethem.Itisuncommontofindaclassifyingtechniquethisiscapableofclusteringtoo.TechniquesfromEvolutionaryComputationmaychangethispractice.Basic

7、allymachinelearningisoneofthelaststepsintheprocessofdatamining.Itisthemomentwherewetransformrawdataintoknowledge,whichmakesitanimportantstepasthisformsthemainideabehinddatamining.Herewefocusonthoseproblemsfromthetwoareasthathavebeensolvedwithevolutionarycomputation.Ashor

8、toverviewofthesefollowsnext.There-afterwegiveanoverviewoftheempiricalresearchwehaveperformedoverthepast