版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领
文档简介
MachineLearning07一月2023Machinelearning,asabranchofartificialintelligence,isgeneraltermsofakindofanalyticalmethod.Itmainlyutilizesputersimulateorrealizethelearnedbehaviorofhuman.07一月202307一月20231)Machinelearningjustlikeatruechampionwhichgohaughtily;
2)Patternrecognitioninprocessofdeclineanddieout;
3)Deeplearningisabrand-newandrapidlyrisingfield.theGooglesearchindexofthreeconceptsince202307一月2023Theconstructedmachinelearningsystembasedonputermainlycontainstwocoreparts:representationandgeneralization.Thefirststepfordatalearningistorepresentthedata,i.e.detectthepatternofdata.Establishageneralizedmodelofdataspaceaccordingtoagroupofknowndatatopredictthenewdata.Thecoretargetofmachinelearningistogeneralizefromknownexperience.Generalizationmeansapowerofwhichthemachinelearningsystemtobelearnedforknowndatathatcouldpredictthenewdata.SupervisedlearningInputdatahaslabels.Themonkindoflearningalgorithmisclassification.Themodelhasbeentrainedviathecorrespondencebetweenfeatureandlabelofinputdata.Therefore,whensomeunknowndatawhichhasfeaturesbutnolabelinput,wecanpredictthelabelofunknowndataaccordingtotheexistingmodel.07一月2023UnsupervisedlearningInputdatahasnolabels.Itrelatestoanotherlearningalgorithm,i.e.clustering.Thebasicdefinitionisacoursethatdividethegatherofphysicalorabstractobjectintomultipleclasswhichconsistofsimilarobjects.07一月2023Iftheoutputeigenvectormarksefromalimitedsetthatconsistofclassornamevariable,thenthekindofmachinelearningbelongstoclassificationproblem.Ifoutputmarkisacontinuousvariable,thenthekindofmachinelearningbelongstoregressionproblem.07一月2023ClassificationstepFeatureextractionFeatureselectionModeltrainingClassificationandpredictionRawdataNewdata07一月2023Featureselection(featurereduction)CurseofDimensionality:Usuallyrefertotheproblemthatconcernedaboutputationofvector.Withtheincreaseofdimension,calculatedamountwilljumpexponentially.Corticalfeaturesofdifferentbrainregionsexhibitvarianteffectduringtheclassificationprocessandmayexistsomeredundantfeature.Inparticularafterthemultimodalfusion,theincreaseoffeaturedimensionwillcause“curseofDimensionality”.07一月2023PrincipalComponentAnalysis,PCAPCAisthemostmonlineardimensionreductionmethod.Itstargetismappingthedataofhighdimensiontolow-dimensionspaceviacertainlinearprojection,andexpectthevarianceofdatathatprojectthecorrespondingdimensionismaximum.Itcanusefewerdatadimensionmeanwhileretainthemajorcharacteristicofrawdata.07一月2023Lineardiscriminantanalysis,LDAThebasicideaofLDAisprojection,mappingtheNdimensiondatatolow-dimensionspaceandseparatethebetween-groupsassoonaspossible.i.e.theoptimalseparabilityinthespace.Thebenchmarkisthenewsubspacehasmaximumbetweenclassdistanceandminimalinter-objectdistance.07一月2023Independentponentanalysis,ICAThebasicideaofICAistoextracttheindependencesignalfromagroupofmixedobservedsignaloruseindependencesignaltorepresentothersignal.07一月2023Recursivefeatureeliminationalgorithm,RFERFEisagreedyalgorithmthatwipeoffinsignificancefeaturestepbysteptoselectthefeature.Firstly,cyclicorderingthefeatureaccordingtotheweightofsub-featureinclassificationandremovethefeaturewhichrankatterminalonebyone.Then,accordingtothefinalfeatureorderinglist,selectdifferentdimensionofseveralfeaturesubsetfronttoback.Assesstheclassificationeffectofdifferentfeaturesubsetandthengettheoptimalfeaturesubset.
07一月2023Classificationalgorithm
DecisiontreeDecisiontreeisatreestructure.Eachnonleafnodeexpressesthetestofafeaturepropertyandeachbranchexpressestheoutputoffeaturepropertyincertainrangeandeachleafnodestoresaclass.Thedecision-makingcourseofdecisiontreeisstartingfromrootnode,testingthecorrespondingfeaturepropertyofwaitingobjects,selectingtheoutputbranchaccordingtotheirvalues,untilreachingtheleafnodeandtaketheclassthatleafnodestoreasthedecisionresult.07一月2023NaiveBayes,NBNBclassificationalgorithmisaclassificationmethodinstatistics.Ituseprobabilitystatisticsknowledgeforclassification.Thisalgorithmcouldapplytolargedatabaseandithashighclassificationaccuracyandhighspeed.07一月2023Artificialneuralnetwork,ANNANNisamathematicalmodelthatapplyakindofstructurewhichsimilarwithsynapseconnectionforinformationprocessing.Inthismodel,amassofnodeformanetwork,i.e.neuralnetwork,toreachthegoalofinformationprocessing.Neuralnetworkusuallyneedtotrain.Thecourseoftrainingisnetworklearning.Thetrainingchangethelinkweightofnetworknodeandmakeitpossessthefunctionofclassification.Thenetworkaftertrainingapplytorecognizeobject.07一月2023k-NearestNeighbors,kNNkNNalgorithmisakindofclassificationmethodbaseonlivingexample.Thismethodistofindthenearestktrainingsampleswithunknownsamplexandexaminethemostofksamplesbelongtowhichclass,thenxbelongstothatclass.kNNisalazylearningmethod.Itstoressamplesbutproceedclassificationuntilneedtoclassify.Ifsamplesetarerelativelyplex,itmaybeleadtolargeputationoverhead.Soitcannotapplytostronglyreal-timeoccasion.07一月2023supportvectormachine,SVMMappingthelinearlyinseparabledatainlow-dimensionspacetohigh-dimensionspaceandmakeitlinearlyseparable07一月2023Crossvalidation,CVThebasicideaofCVisgroupingtherawdatainasense.Onepartistakenastrainset,theotherpartistakenasvalidationset.Primarily,theclassifieristrainedwithtrainset,andthenusevalidationsettotestthereceivedmodelbytraining.07一月2023K-foldcross-validationIn
k-foldcross-validation,theoriginalsampleisrandomlypartitionedinto
k
equalsizedsubsamples.Ofthe
k
subsamples,asinglesubsampleisretainedasthevalidationdatafortestingthemodel,andtheremaining
k
−
1subsamplesareusedastrainingdata.Thecross-validationprocessisthenrepeated
k
times(the
folds),witheachofthe
k
subsamplesusedexactlyonceasthevalidationdata.The
k
resultsfromthefoldscanthenbeaveragedtoproduceasingleestimation.Theadvantageofthismethodoverrepeatedrandomsub-samplingisthatallobservationsareusedforbothtrainingandvalidation,andeachobservationisusedforvalidationexactlyonce.10-foldcross-validationismonlyused.07一月2023Leave-one-outcross-validation,LOOCVWhen
k
=
n
(thenumberofobservations),the
k-foldcross-validationisexactlytheleave-one-outcross-validation.07一月2023confusionmatrixTP——goldstandardandtestaffirmsufferfromcertainillness;TN——goldstan
温馨提示
- 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
- 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
- 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
- 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
- 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
- 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
- 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。
最新文档
- 2025 小学四年级科学下册水蒸发快慢实验变量控制表课件
- 2026年大公司战略管理研究试题
- 2026年电商客户服务专家高级模拟测试题
- 2026年初级音乐教师招聘笔试模拟题
- 2026年化工厂紧急泄险措施安全保障试题设计
- 2026年环境保护与可持续发展知识试题
- 2026年软件工程专业知识试题软件工程原理与软件开发流程考试题
- 2026年提升职业技能工程安全与卫生职称评审实务试题集
- 保密协议2026年保密培训要求
- 水电站辅助设施建设方案
- 《筑牢安全防线 欢度平安寒假》2026年寒假安全教育主题班会课件
- 养老院老人生活设施管理制度
- 2026年税务稽查岗位考试试题及稽查实操指引含答案
- (2025年)林业系统事业单位招聘考试《林业知识》真题库与答案
- 2026版安全隐患排查治理
- 道路施工安全管理课件
- 2026年七台河职业学院高职单招职业适应性考试备考题库有答案解析
- 办公楼电梯间卫生管理方案
- 新生儿休克诊疗指南
- 专题学习活动 期末复习课件 新教材统编版八年级语文上册
- VTE患者并发症预防与处理
评论
0/150
提交评论