资源描述：
《00 - Editorial. Human and Machine Learning.pdf》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、MachineLearning1:243-248,1986©1986KluwerAcademicPublishers,Boston-ManufacturedinTheNetherlandsEditorial:HumanandMachineLearningThegoalsofmachinelearningOnecanidentifyanumberofdifferentthemeswithinthemachinelearningcommunity,eachcorrespondingtocentralgoalsofitsparentfield,artificialintelligence.Fo
2、rinstance,manyAIresearchersareconcernedwithimplementingknowledge-intensivesystems,buttheseoftentakeman-yearstoconstruct.Machinelearningmayprovidemethodsforautomatingthisprocess,promisingconsider-ablesavingsintimeandeffort.Similarly,manyAIresearchersviewartificialintelligenceasascientificdisciplin
3、eratherthananengineeringone,andhopetoformulategeneralprinciplesofintelligentbehaviorthatholdacrossavarietyofdomains.Sincemachinelearningfocusesontheacquisitionofdomain-specificknowledgeratherthantheknowledgeitself,itholdsconsiderablepotentialforsuchgeneralprinciples.StillotherAIresearchersareconc
4、ernedwithexplaininghumanbehavior,viewingAItechniquesasanidealtoolforstatingcomplextheoriesofthehumaninformationprocessingsystem.Sincelearningisacentralphenomenoninhumancognition,theseresearchersevaluatemachinelearningmethodsintermsoftheirabilitytoexplainhumanlearning.Althoughthisisaminorityopinio
5、nwithinmachinelearning,belowwepresentsomereasonswhymoremembersofthefieldshouldtakethisapproachseriously.ScienceassearchOneofthecentralinsightsofAIisthatintelligenceinvolvessearch,andthateffectivesearchisconstrainedbydomain-specificknowledge.Thisframeworkcanbeappliedtoproblemsolving,languageunders
6、tanding,andlearningfromexperience.Onecanevenapplythissearchmetaphortomachinelearningasafieldofscientificstudy.Inthisframework,machinelearningresearchersareexploringavastspaceofpossiblelearningmethods,searchingfortechniqueswithusefulcharacteristicsandlookingforrelationsbetweenthesemethods.Expandin
7、gonthisanalogy,anydomain-specificconstraintsthatwecanplaceonthissearchshould,easeourtaskconsiderably.Inparticular,ourknowledgeof244P.LANGLEYhumanlearningprovidesanimportantsetofconstraintsonourmodelsoflearning.Someofth
