ActiveLearningforProbabilisticStructuredPredictionofCutsandMatchingsSimaBehpour12AnqiLiu3BrianD.Ziebart2AbstractFigure1.Anexampleofabipartitematchinginavideotrackingapplication.Whensomeassignmentsa...
AccelerationofSVRGandKatyushaXbyInexactPreconditioningYanliLiu1FeiFeng1WotaoYin1Abstractregularizerψ(x)isproper,closed,andconvex,butmaybenonsmooth.Anonzeroψ(x)isdesirableinmanyapplica-Empiricalri...
AcceleratedLinearConvergenceofStochasticMomentumMethodsinWassersteinDistancesBugraCan1MertGurbuzbalaban1LingjiongZhu2Abstractsupervisedlearningincludelinearandnon-linearregressionproblems,supportve...
ATree-BasedMethodforFastRepeatedSamplingofDeterminantalPointProcessesJenniferGillenwater1AlexKulesza1ZeldaMariet23SergeiVassilvitskii1AbstractForrecommendersystems,diversityintroducesvarietyandincr...
ATheoryofRegularizedMarkovDecisionProcessesMatthieuGeist1BrunoScherrer2OlivierPietquin1AbstractTsallisentropy(Leeetal.,2018),withthemotivationofhavingasparseregularizedgreedypolicy.OtherapproachesM...
ATheoreticalAnalysisofContrastiveUnsupervisedRepresentationLearningSanjeevArora12HrishikeshKhandeparkar1MikhailKhodak3OrestisPlevrakis1NikunjSaunshi1AbstractForimages,aproofofexistenceforbroadlyuse...
ATail-IndexAnalysisofStochasticGradientNoiseinDeepNeuralNetworksUmutS¸ims¸ekli1LeventSagun2MertGu¨rbu¨zbalaban3Abstractmanyapplicationdomains(LeCunetal.,2015;Krizhevskyetal.,2012;Hintonetal.,20...
AStatisticalInvestigationofLongMemoryinLanguageandMusicAlexanderGreaves-Tunnell1ZaidHarchaoui1Abstractundoubtedlyhelpful,suchheuristicsarerarelydefinedwithrespecttoanunderlyingmathematicalorstatist...
AQuantitativeAnalysisoftheEffectofBatchNormalizationonGradientDescentYongqiangCai1QianxiaoLi12ZuoweiShen1AbstracteffectsofBNareattributedtotheso-called“reductionofcovariateshift”.However,itisuncl...
AKernelTheoryofModernDataAugmentationTriDao1AlbertGu1AlexanderJ.Ratner1VirginiaSmith2ChristopherDeSa3ChristopherRe´1Abstractasregularizertomaketheresultingmodelmorerobust,andprovideresourcestodata...
WhatistheEffectofImportanceWeightinginDeepLearning?JonathonByrd1ZacharyC.Lipton1AbstractEq[f(x)],importancesamplingproducesanunbiasedesti-matebyweightingeachsamplexaccordingtothelikelihoodImportanc...
WeakDetectionofSignalintheSpikedWignerModelHyeWonChung1JiOonLee2Abstractwherethesignalx∈RNandHisanN×NWignerma-trix.(SeeDefinitions1and2.)ThespikedWignermodelWeconsidertheproblemofdetectingthepres...
WassersteinofWassersteinLossforLearningGenerativeModelsYonatanDukler1WuchenLi1AlexTongLin1GuidoMontu´far123AbstractGenerativeAdversarialNetworks(GANs).TheapplicationoftheWassersteinmetrictodefinet...
TowardsFastComputationofCertifiedRobustnessforReLUNetworksTsui-WeiWeng1HuanZhang2HonggeChen1ZhaoSong34Cho-JuiHsieh2DuaneBoning1InderjitS.Dhillon4LucaDaniel1Abstract1.IntroductionVerifyingtherobustn...
TheoreticalAnalysisofSparseSubspaceClusteringwithMissingEntriesManolisC.Tsakiris1Rene´Vidal2Abstractfoundnumerousapplicationsinmachinelearning,com-putervision,patternrecognition,bioinformaticsands...
TheoreticalAnalysisofImage-to-ImageTranslationwithAdversarialLearningXudongPan1MiZhang1DaizongDing1Abstractagenerator(i.e.anadaptivemodelthatmapsagaussiannoisetoafakesample)andadiscriminator(i.e.an...
TheMultilinearStructureofReLUNetworksThomasLaurent1JamesH.vonBrecht2AbstractΩ2Ω3WestudythelosssurfaceofneuralnetworksΩ1Ω4equippedwithahingelosscriterionandReLUorleakyReLUnonlinearities.Anysuchn...
ThePowerofInterpolation:UnderstandingtheEffectivenessofSGDinModernOver-parametrizedLearning†SiyuanMa1RaefBassily1MikhailBelkin1Abstract1IntroductionInthispaperweaimtoformallyexplainthephe-Mostmach...
TheMechanicsofn-PlayerDifferentiableGamesDavidBalduzzi1Se´bastienRacanie`re1JamesMartens1JakobFoerster2KarlTuyls1ThoreGraepel1Abstractoptimization(Pfau&Vinyals,2016),syntheticgradients(Jaderberget...
TheLimitsofMaxing,Ranking,andPreferenceLearningMoeinFalahatgar1AyushJain1AlonOrlitsky1VenkatadheerajPichapati1VaishakhRavindrakumar1Abstractthatguaranteestheexistenceofmaximumandshowthatunderthismo...