![機(jī)器學(xué)習(xí)的幾何觀點(diǎn)-LAMDA課件_第1頁(yè)](http://file4.renrendoc.com/view/218e3f9d05dccf1c867ee7a722bd388e/218e3f9d05dccf1c867ee7a722bd388e1.gif)
![機(jī)器學(xué)習(xí)的幾何觀點(diǎn)-LAMDA課件_第2頁(yè)](http://file4.renrendoc.com/view/218e3f9d05dccf1c867ee7a722bd388e/218e3f9d05dccf1c867ee7a722bd388e2.gif)
![機(jī)器學(xué)習(xí)的幾何觀點(diǎn)-LAMDA課件_第3頁(yè)](http://file4.renrendoc.com/view/218e3f9d05dccf1c867ee7a722bd388e/218e3f9d05dccf1c867ee7a722bd388e3.gif)
![機(jī)器學(xué)習(xí)的幾何觀點(diǎn)-LAMDA課件_第4頁(yè)](http://file4.renrendoc.com/view/218e3f9d05dccf1c867ee7a722bd388e/218e3f9d05dccf1c867ee7a722bd388e4.gif)
![機(jī)器學(xué)習(xí)的幾何觀點(diǎn)-LAMDA課件_第5頁(yè)](http://file4.renrendoc.com/view/218e3f9d05dccf1c867ee7a722bd388e/218e3f9d05dccf1c867ee7a722bd388e5.gif)
版權(quán)說(shuō)明:本文檔由用戶(hù)提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請(qǐng)進(jìn)行舉報(bào)或認(rèn)領(lǐng)
文檔簡(jiǎn)介
AGeometricPerspectiveonMachineLearning何曉飛浙江大學(xué)計(jì)算機(jī)學(xué)院1AGeometricPerspectiveonMacMachineLearning:theproblemf何曉飛Information(trainingdata)
f:X→YXandYareusuallyconsideredasaEuclideanspaces.2MachineLearning:theproblemfManifoldLearning:geometricperspectiveThedataspacemaynotbeaEuclideanspace,butanonlinearmanifold.?
Euclideandistance.?
fisdefinedonEuclideanspace.?ambientdimension? geodesicdistance.?fisdefinedonnonlinearmanifold.?manifolddimension.instead…3ManifoldLearning:geometricpManifoldLearning:thechallengesThemanifoldisunknown!Wehaveonlysamples!HowdoweknowMisasphereoratorus,orelse?HowtocomputethedistanceonM?
versusThisisunknown:Thisiswhatwehave:??orelse…?TopologyGeometryFunctionalanalysis4ManifoldLearning:thechallenManifoldLearning:currentsolutionFindaEuclideanembedding,andthenperformtraditionallearningalgorithmsintheEuclideanspace.5ManifoldLearning:currentsolSimplicity6Simplicity6Simplicity7Simplicity7Simplicityisrelative8Simplicityisrelative8Manifold-basedDimensionalityReductionGivenhighdimensionaldatasampledfromalowdimensionalmanifold,howtocomputeafaithfulembedding?Howtofindthemappingfunction?Howtoefficientlyfindtheprojectivefunction?9Manifold-basedDimensionalityAGoodMappingFunctionIfxiandxjareclosetoeachother,wehopef(xi)andf(xj)preservethelocalstructure(distance,similarity…)k-nearestneighborgraph:Objectivefunction:Differentalgorithmshavedifferentconcerns10AGoodMappingFunctionIfxiLocalityPreservingProjectionsPrinciple:ifxiandxjareclose,thentheirmapsyiandyjarealsoclose.11LocalityPreservingProjectionLocalityPreservingProjectionsPrinciple:ifxiandxjareclose,thentheirmapsyiandyjarealsoclose.Mathematicalformulation:minimizetheintegralofthegradientoff.12LocalityPreservingProjectionLocalityPreservingProjectionsPrinciple:ifxiandxjareclose,thentheirmapsyiandyjarealsoclose.Mathematicalformulation:minimizetheintegralofthegradientoff.Stokes’Theorem:13LocalityPreservingProjectionLocalityPreservingProjectionsPrinciple:ifxiandxjareclose,thentheirmapsyiandyjarealsoclose.Mathematicalformulation:minimizetheintegralofthegradientoff.Stokes’Theorem:LPPfindsalinearapproximationtononlinearmanifold,whilepreservingthelocalgeometricstructure.14LocalityPreservingProjectionManifoldofFaceImagesExpression(Sad>>>Happy)
Pose(Right>>>Left)15ManifoldofFaceImagesExpressManifoldofHandwrittenDigitsThicknessSlant16ManifoldofHandwrittenDigitsLearningtarget:TrainingExamples:LinearRegressionModelActiveandSemi-SupervisedLearning:AGeometricPerspective17Learningtarget:ActiveandSemGeneralizationErrorGoalofRegression
Obtainalearnedfunctionthatminimizesthegeneralizationerror(expectederrorforunseentestinputpoints).MaximumLikelihoodEstimate18GeneralizationErrorGoalofReGauss-MarkovTheoremForagivenx,theexpectedpredictionerroris:19Gauss-MarkovTheoremForagiveGauss-MarkovTheoremForagivenx,theexpectedpredictionerroris:Good!Bad!20Gauss-MarkovTheoremForagiveExperimentalDesignMethodsThreemostcommonscalarmeasuresofthesizeoftheparameter(w)covariancematrix:A-optimalDesign:determinantofCov(w).D-optimalDesign:traceofCov(w).E-optimalDesign:maximumeigenvalueofCov(w).Disadvantage:thesemethodsfailtotakeintoaccountunmeasured(unlabeled)datapoints.21ExperimentalDesignMethodsThrManifoldRegularization:Semi-SupervisedSettingMeasured(labeled)points:discriminantstructureUnmeasured(unlabeled)points:geometricalstructure?22ManifoldRegularization:Semi-Measured(labeled)points:discriminantstructureUnmeasured(unlabeled)points:geometricalstructure?randomlabelingManifoldRegularization:Semi-SupervisedSetting23Measured(labeled)points:disMeasured(labeled)points:discriminantstructureUnmeasured(unlabeled)points:geometricalstructure?randomlabelingactivelearningactivelearning+semi-supervsedlearningManifoldRegularization:Semi-SupervisedSetting24Measured(labeled)points:disUnlabeledDatatoEstimateGeometryMeasured(labeled)points:discriminantstructure25UnlabeledDatatoEstimateGeoUnlabeledDatatoEstimateGeometryMeasured(labeled)points:discriminantstructureUnmeasured(unlabeled)points:geometricalstructure26UnlabeledDatatoEstimateGeoUnlabeledDatatoEstimateGeometryMeasured(labeled)points:discriminantstructureUnmeasured(unlabeled)points:geometricalstructureComputenearestneighborgraphG27UnlabeledDatatoEstimateGeoUnlabeledDatatoEstimateGeometryMeasured(labeled)points:discriminantstructureUnmeasured(unlabeled)points:geometricalstructureComputenearestneighborgraphG28UnlabeledDatatoEstimateGeoUnlabeledDatatoEstimateGeometryMeasured(labeled)points:discriminantstructureUnmeasured(unlabeled)points:geometricalstructureComputenearestneighborgraphG29UnlabeledDatatoEstimateGeoUnlabeledDatatoEstimateGeometryMeasured(labeled)points:discriminantstructureUnmeasured(unlabeled)points:geometricalstructureComputenearestneighborgraphG30UnlabeledDatatoEstimateGeoUnlabeledDatatoEstimateGeometryMeasured(labeled)points:discriminantstructureUnmeasured(unlabeled)points:geometricalstructureComputenearestneighborgraphG31UnlabeledDatatoEstimateGeoLaplacianRegularizedLeastSquare(BelkinandNiyogi,2006)LinearobjectivefunctionSolution32LaplacianRegularizedLeastSqActiveLearningHowtofindthemostrepresentativepointsonthemanifold?33ActiveLearningHowtofindtheObjective:Guidetheselectionofthesubsetofdatapointsthatgivesthemostamountofinformation.Experimentaldesign:selectsamplestolabelManifoldRegularizedExperimentalDesignSharethesameobjectivefunctionasLaplacianRegularizedLeastSquares,simultaneouslyminimizetheleastsquareerroronthemeasuredsamplesandpreservethelocalgeometricalstructureofthedataspace.ActiveLearning34Objective:Guidetheselection
,Inordertomaketheestimatorasstableaspossible,thesizeofthecovariancematrixshouldbeassmallaspossible.D-optimality:minimizethedeterminantofthecovariancematrixAnalysisofBiasandVariance35
Selectthefirstdatapointsuchthatismaximized,Supposekpointshavebeenselected,choosethe(k+1)thpointsuchthat.UpdateManifoldRegularizedExperimentalDesignWhereareselectedfromThealgorithm36ManifoldRegularizedExperimenConsiderfeaturespaceFinducedbysomenonlinearmappingφ,and<f(xi),f(xj)>=K(xi,xi).K(·,·):positivesemi-definitekernelfunctionRegressionmodelinRKHS:ObjectivefunctioninRKHS:NonlinearGeneralizationinRKHS37ConsiderfeaturespaceFinducSelectthefirstdatapointsuchthatismaximized,Supposekpointshavebeenselected,choosethe(k+1)thpointsuchthat.UpdateKernelGraphRegularizedExperimentalDesignwhereareselectedfromNonlinearGeneralizationinRKHS38KernelGraphRegularizedExperASyntheticExampleA-optimalDesignLaplacianRegularizedOptimalDesign39ASyntheticExampleA-optimalDASyntheticExampleA-optimalDesignLaplacianRegularizedOptima
溫馨提示
- 1. 本站所有資源如無(wú)特殊說(shuō)明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請(qǐng)下載最新的WinRAR軟件解壓。
- 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請(qǐng)聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶(hù)所有。
- 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁(yè)內(nèi)容里面會(huì)有圖紙預(yù)覽,若沒(méi)有圖紙預(yù)覽就沒(méi)有圖紙。
- 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
- 5. 人人文庫(kù)網(wǎng)僅提供信息存儲(chǔ)空間,僅對(duì)用戶(hù)上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對(duì)用戶(hù)上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對(duì)任何下載內(nèi)容負(fù)責(zé)。
- 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請(qǐng)與我們聯(lián)系,我們立即糾正。
- 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶(hù)因使用這些下載資源對(duì)自己和他人造成任何形式的傷害或損失。
最新文檔
- 體育場(chǎng)地施工法律法規(guī)解讀考核試卷
- 樂(lè)器批發(fā)商的顧客關(guān)系管理提升路徑考核試卷
- 光伏開(kāi)發(fā)合同范本
- 未來(lái)展覽空間的環(huán)境設(shè)計(jì)與技術(shù)實(shí)現(xiàn)案例研究
- 工程監(jiān)理工作方法及措施
- 技術(shù)服務(wù)財(cái)務(wù)分析與決策支持考核試卷
- 中介工資合同范本
- 修公路勞務(wù)合同范本
- 買(mǎi)房按揭買(mǎi)賣(mài)合同范本
- 2025-2030年新能源汽車(chē)文化傳播行業(yè)跨境出海戰(zhàn)略研究報(bào)告
- 2024政府采購(gòu)評(píng)審專(zhuān)家考試真題庫(kù)及答案
- 2025年道路貨運(yùn)駕駛員從業(yè)資格證模擬考試題
- 數(shù)學(xué)-安徽省皖南八校2025屆高三上學(xué)期12月第二次大聯(lián)考試題和答案
- 退市新規(guī)解讀-上海證券交易所、大同證券
- 融資報(bào)告范文模板
- 桃李面包盈利能力探析案例11000字
- GB/Z 30966.71-2024風(fēng)能發(fā)電系統(tǒng)風(fēng)力發(fā)電場(chǎng)監(jiān)控系統(tǒng)通信第71部分:配置描述語(yǔ)言
- 腦梗死的護(hù)理查房
- 2025高考數(shù)學(xué)專(zhuān)項(xiàng)復(fù)習(xí):概率與統(tǒng)計(jì)的綜合應(yīng)用(十八大題型)含答案
- 產(chǎn)后抑郁癥講課課件
- 2024-2030年中國(guó)紫蘇市場(chǎng)深度局勢(shì)分析及未來(lái)5發(fā)展趨勢(shì)報(bào)告
評(píng)論
0/150
提交評(píng)論