机器学习介绍(英文版：备注里有中文翻译).ppt

《机器学习介绍(英文版：备注里有中文翻译).ppt》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、MachineLearning20九月2021Machinelearning,asabranchofartificialintelligence,isgeneraltermsofakindofanalyticalmethod.Itmainlyutilizescomputersimulateorrealizethelearnedbehaviorofhuman.20九月202120九月20211）Machinelearningjustlikeatruechampionwhichgohaughtily;2）Patternrecognitioninproce

2、ssofdeclineanddieout;3）Deeplearningisabrand-newandrapidlyrisingfield.theGooglesearchindexofthreeconceptsince200420九月2021Theconstructedmachinelearningsystembasedoncomputermainlycontainstwocoreparts：representationandgeneralization.Thefirststepfordatalearningistorepresentthedata,i

3、.e.detectthepatternofdata.Establishageneralizedmodelofdataspaceaccordingtoagroupofknowndatatopredictthenewdata.Thecoretargetofmachinelearningistogeneralizefromknownexperience.Generalizationmeansapowerofwhichthemachinelearningsystemtobelearnedforknowndatathatcouldpredictthenewda

4、ta.SupervisedlearningInputdatahaslabels.Thecommonkindoflearningalgorithmisclassification.Themodelhasbeentrainedviathecorrespondencebetweenfeatureandlabelofinputdata.Therefore,whensomeunknowndatawhichhasfeaturesbutnolabelinput,wecanpredictthelabelofunknowndataaccordingtotheexist

5、ingmodel.20九月2021UnsupervisedlearningInputdatahasnolabels.Itrelatestoanotherlearningalgorithm,i.e.clustering.Thebasicdefinitionisacoursethatdividethegatherofphysicalorabstractobjectintomultipleclasswhichconsistofsimilarobjects.20九月2021Iftheoutputeigenvectormarkscomefromalimited

6、setthatconsistofclassornamevariable,thenthekindofmachinelearningbelongstoclassificationproblem.Ifoutputmarkisacontinuousvariable,thenthekindofmachinelearningbelongstoregressionproblem.20九月2021ClassificationstepFeatureextractionFeatureselectionModeltrainingClassificationandpredi

7、ctionRawdataNewdata20九月2021Featureselection（featurereduction）CurseofDimensionality：Usuallyrefertotheproblemthatconcernedaboutcomputationofvector.Withtheincreaseofdimension,calculatedamountwilljumpexponentially.Corticalfeaturesofdifferentbrainregionsexhibitvarianteffectduringthe

8、classificationprocessandmayexistsomeredundantfeature.I