ProvableMeta-LearningofLinearRepresentationsNileshTripuraneni1ChiJin2MichaelI.Jordan1Abstractrangingfromdeepreinforcementlearning(Baevskietal.,2019)tonaturallanguageprocessing(Ando&Zhang,2005;Meta-...
ProvableRobustnessofAdversarialTrainingforLearningHalfspaceswithNoiseDifanZou1SpencerFrei2QuanquanGu1AbstractToformalizetheabovecomment,letusdefinethero-Weanalyzethepropertiesofadversarialtrain-bus...
ProvableLipschitzCertificationforGenerativeModelsMattJordan1AlexandrosG.Dimakis1Abstractthroughavarietyofnonlinearoperatorsandrelatethistoa2-dimensionalgeometricproblemthatwesolveop-Wepresentascala...
ProvableGeneralizationofSGD-trainedNeuralNetworksofAnyWidthinthePresenceofAdversarialLabelNoiseSpencerFrei1YuanCao2QuanquanGu2AbstractdefineWeconsideraone-hidden-layerleakyReLUnet-OPT:=minP(x,y)∼D...
Model-TargetedPoisoningAttackswithProvableConvergenceFnuSuya1SaeedMahloujifar2AnshumanSuri1DavidEvans1YuanTian1AbstractMostworkonpoisoningattackshasconsideredoneoftwoextremalattackerobjectives:indi...
CrowdsourcingviaAnnotatorCo-occurrenceImputationandProvableSymmetricNonnegativeMatrixFactorizationShahanaIbrahim1XiaoFu1Abstractannotationrequiressubstantialhumaneffortswithconsider-ableexpertise,w...
BilinearClasses:AStructuralFrameworkforProvableGeneralizationinRLSimonS.Du1ShamM.Kakade1JasonD.Lee2ShacharLovett3GauravMahajan3WenSun4RuosongWang5AbstractthereisalsoarealizationthatpracticalRLappro...
AModularAnalysisofProvableAccelerationviaPolyak’sMomentum:TrainingaWideReLUNetworkandaDeepLinearNetworkJun-KunWang1Chi-HengLin2JacobAbernethy1Abstract1.IntroductionIncorporatingaso-called“momentu...
Second-OrderProvableDefensesagainstAdversarialAttacksSahilSingla1SoheilFeizi1Abstract1.IntroductionArobustnesscertificateagainstadversarialexam-Modernneuralnetworksachievehighaccuracyontasksplesist...
RandomHypervolumeScalarizationsforProvableMulti-ObjectiveBlackBoxOptimizationDanielGolovin1Qiuyi(Richard)Zhang1Abstractcatedlossesusedindeeplearning(LeCunetal.,2015)andreinforcementlearning(Suttone...
ProvableSelf-PlayAlgorithmsforCompetitiveReinforcementLearningYuBai1ChiJin2Abstractconflictingrewards(sothattheyessentiallycompetewitheachother)yetcanbetrainedinacentralizedfashion(i.e.Self-play,wh...
ProvableSmoothnessGuaranteesforBlack-BoxVariationalInferenceJustinDomke1AbstractRanganathetal.,2014;Regieretal.,2017a;Kucukelbiretal.,2017)hasfocusedona“blackbox”modelwhereBlack-boxvariationalinf...
ProvableRepresentationLearningforImitationLearningviaBi-levelOptimizationSanjeevArora12SimonS.Du2ShamKakade3YupingLuo1NikunjSaunshi1AbstractMarkovdecisionprocesses(MDPs)thatsharethesamestateandacti...
ProvableGuaranteesforDecisionTreeInduction:TheAgnosticSettingGuyBlanc1JaneLange1Li-YangTan1AbstractnowthattheybuildadecisiontreeTforabinaryclassifierf:Rn→{0,1}inagreedy,top-downfashion:Wegivestren...
FromSetstoMultisets:ProvableVariationalInferenceforProbabilisticIntegerSubmodularModelsAytuncSahin1YataoBian2JoachimM.Buhmann1AndreasKrause1Abstracttionperspective.Usingthecelebratedgreedyalgorithm...
BreakingtheCurseofManyAgents:ProvableMeanEmbeddingQ-IterationforMean-FieldReinforcementLearningLingxiaoWang1ZhuoranYang2ZhaoranWang1Abstract1998;Dzˇeroskietal.,2001;Guestrinetal.,2002;Karetal.,201...
ProvableGuaranteesforGradient-BasedMeta-LearningMikhailKhodak1Maria-FlorinaBalcan1AmeetTalwalkar12Abstractwithexamples(xi,yi)∈X×Yforanunseentask,thepopularMAMLalgorithm(Finnetal.,2017)outputsWest...
ProvableDefensesagainstAdversarialExamplesviatheConvexOuterAdversarialPolytopeEricWong1J.ZicoKolter2Abstract1.IntroductionWeproposeamethodtolearndeepReLU-basedRecentworkindeeplearninghasdemonstrate...
ProvableVariableSelectionforStreamingFeaturesJingWang1JieShen2PingLi3Abstractandonlyrelatedfeatureswillbeincludedintherecommen-dationmodel.Hence,inreal-worldapplications,featuresInlarge-scalemachin...
ProvableAlternatingGradientDescentforNon-negativeMatrixFactorizationwithStrongCorrelationsYuanzhiLi1YingyuLiang1AbstractBydoingso,onecanavoidcancellationofdifferentfea-turesandimproveinterpretabili...