UnderstandingtheDynamicsofGradientFlowinOverparameterizedLinearModelsSalmaTarmoun12GuilhermeFranc¸a13BenjaminHaeffele14Rene´Vidal14Abstractathoroughanalysisofthedynamicsofoptimizationmeth-ods,suc...
UnderstandingNoiseInjectioninGANsRuiliFeng1DeliZhao2Zheng-JunZha†1Abstractquantizingneuralnetworks(Baskinetal.,2018).Inrecentyears,noiseinjection1hasattractedmoreandmoreatten-Noiseinjectionisaneff...
UnderstandingInvarianceviaFeedforwardInversionofDiscriminativelyTrainedClassifiersPiotrTeterwak1ChiyuanZhang2DilipKrishnan2MichaelC.Mozer23Abstract1.IntroductionAdiscriminativelytrainedneuralnetcla...
UnderstandingSelf-SupervisedLearningDynamicswithoutContrastivePairsYuandongTian1XinleiChen1SuryaGanguli12Abstractquiringexpensivetargetlabels(Devlinetal.,2018).Manystate-of-the-artSSLmethodsincompu...
UnderstandingInstance-LevelLabelNoise:DisparateImpactsandTreatmentsYangLiu1Abstractframework(Feldman,2020),wedemonstratethevaryingeffectsofmemorizingnoisylabelsassociatedwithinstancesThispaperaimst...
UnderstandingFailuresinOut-of-DistributionDetectionwithDeepGenerativeModelsLilyH.Zhang1MarkGoldstein1RajeshRanganath1Abstractdistribution,estimatedviaamodel(Bishop,1994).Deepgenerativemodels(DGMs)s...
UnderstandingandMitigatingAccuracyDisparityinRegressionJianfengChi1YuanTian1GeoffreyJ.Gordon2HanZhao3Abstractcases,BuolamwiniandGebru(2018)showedthatnotableaccuracydisparityexistsacrossdifferentrac...
TowardsUnderstandingLearninginNeuralNetworkswithLinearTeachersRoeiSarussi1AlonBrutzkus1AmirGloberson1Abstractwillseparatethedata.Whichofthesewillbefoundbygradientdescent?Cananeuralnetworkminimizing...
TowardsUnderstandingandMitigatingSocialBiasesinLanguageModelsPaulPuLiang1ChiyuWu1Louis-PhilippeMorency1RuslanSalakhutdinov1AbstractZhaoetal.,2017).Morerecently,languagemodels(LMs)areincreasinglyuse...
TowardUnderstandingtheFeatureLearningProcessofSelf-supervisedContrastiveLearningZixinWen1YuanzhiLi2Abstractcanevenoutperformthoselearnedbysupervisedlearninginseveraldownstreamtasks.Theremakablepote...
QuantitativeUnderstandingofVAEasaNon-linearlyScaledIsometricEmbeddingAkiraNakagawa1KeizoKato1TaijiSuzuki23Abstractconvergestoanappropriatevalue,andsoon.Alemietal.(2018)introducedtheRDtrade-offbased...
BeyondVarianceReduction:UnderstandingtheTrueImpactofBaselinesonPolicyOptimizationWesleyChung1ValentinThomas2MarlosC.Machado345NicolasLeRoux612Abstract&Cesa-Bianchi,2012).Whiletheformermeasureisofte...
UnravelingMeta-Learning:UnderstandingFeatureRepresentationsforFew-ShotTasksMicahGoldblum1StevenReich1LiamFowl1RenkunNi1ValeriiaCherepanova1TomGoldstein1Abstractdownentirelywhenonlyafewdatasamplesar...
UnderstandingSelf-TrainingforGradualDomainAdaptationAnanyaKumar1TengyuMa1PercyLiang1AbstractFigure1.Ingradualdomainadaptationwearegivenlabeleddatafromasourcedomain,andunlabeleddatafromintermediated...
UnderstandingtheImpactofModelIncoherenceonConvergenceofIncrementalSGDwithRandomReshuffeShaocongMa1YiZhou1Abstractthesamplelossofthei-thdatasample.Suchaproblemformulationcoversavarietyofmachinelearn...
UnderstandingtheCurseofHorizoninOff-PolicyEvaluationviaConditionalImportanceSamplingYaoLiu1Pierre-LucBacon2EmmaBrunskill1Abstractincreasinginterestindevelopingaccurateandefficientalgo-rithmsforoff-...
UnderstandingContrastiveRepresentationLearningthroughAlignmentandUniformityontheHypersphereTongzhouWang1PhillipIsola1AbstractContrastiverepresentationlearninghasbeenout-Alignment:Similarsampleshave...
UnderstandingandStabilizingGANs’TrainingDynamicsusingControlTheoryKunXu1ChongxuanLi1JunZhu1BoZhang1Abstractplesfromfakeones.Gaimstorecoverthedatadistributionbymaximizingthisratio.Thisframeworkisfo...
UnderstandingandMitigatingtheTradeoffBetweenRobustnessandAccuracyAditiRaghunathan1SangMichaelXie1FannyYang2JohnC.Duchi1PercyLiang1AbstractStdErr(AT)-StdErr(Std)(%)ε=1/255Figure1.Gapbetweenthestan-...
EstimatingGeneralizationunderDistributionShiftsviaDomain-InvariantRepresentationsChing-YaoChuang1AntonioTorralba1StefanieJegelka1AbstractAISafety(Amodeietal.,2016),motivatingrecentworkonestimatingt...