DessiLBI:ExploringStructuralSparsityofDeepNetworksviaDifferentialInclusionPathsYanweiFu1ChenLiu1DonghaoLi12XinweiSun3JinshanZeng24YuanYao2Abstract1IntroductionOver-parameterizationisubiquitousnowad...
DissectingNon-VacuousGeneralizationBoundsbasedontheMean-FieldApproximationKonstantinosPitas1Abstract(a)Explaininghowoverparametrizedneuralnet-Figure1.Risk-ComplexityplotforMNIST10:Theareabelowworks...
CurseofDimensionalityonRandomizedSmoothingforCertifiableRobustnessAounonKumar1AlexanderLevine1TomGoldstein1SoheilFeizi1AbstractbasedmethodssuchasFGSM(Goodfellowetal.,2015)andprojectedgradientdescen...
ContrastiveMulti-ViewRepresentationLearningonGraphsKavehHassani1AmirHoseinKhasahmadi12Abstractworks(Kipf&Welling,2017),molecules(Duvenaudetal.,2015),andknowledgegraphs(Vivona&Hassani,2019).Weintrod...
BeyondSyntheticNoise:DeepLearningonControlledNoisyLabelsLuJiang1DiHuang2MasonLiu3WeilongYang1AbstractHowever,duetothelackofsuitabledatasets,previousworkhasonlyexaminedDNNsoncontrolledsyntheticlabel...
AnOptimisticPerspectiveonOfflineReinforcementLearningRishabhAgarwal1DaleSchuurmans12MohammadNorouzi1Abstractunsafe,orrequireahigh-fidelitysimulatorthatisoftendiffi-culttobuild(Dulac-Arnoldetal.,201...
AdversarialAttacksonProbabilisticAutoregressiveForecastingModelsRaphae¨lDang-Nhu1GagandeepSingh1PavolBielik1MartinVechev1Abstractsinglebestvaluehasseveraladvantages–itnaturallyfitstheinherentlyst...
AdversarialAttacksonCopyrightDetectionSystemsParsaSaadatpanah1AliShafahi1TomGoldstein1Abstractdetectionextractfeatures,calledfingerprints,fromsam-pledvideooraudio,andthenmatchthesefeatureswithaItis...
ActiveLearningonAttributedGraphsviaGraphCognizantLogisticRegressionandPreemptiveQueryGenerationFlorenceRegol†1SoumyasundarPal1YingxueZhang2MarkCoates1Abstract1.IntroductionNodeclassificationinattr...
ADistributionalViewonMulti-ObjectivePolicyOptimizationAbbasAbdolmaleki1SandyH.Huang1LeonardHasenclever1MichaelNeunert1H.FrancisSong1MartinaZambelli1MuriloF.Martins1NicolasHeess1RaiaHadsell1MartinRi...
UnderstandingtheImpactofEntropyonPolicyOptimizationZafaraliAhmed12NicolasLeRoux13MohammadNorouzi3DaleSchuurmans34Abstractlis,2000;Greensmithetal.,2004;Schulmanetal.,2015b;Tuckeretal.,2018).Entropyr...
UnderstandingMCMCDynamicsasFlowsontheWassersteinSpaceChangLiu1JingweiZhuo1JunZhu1AbstractdistributionminimizestheKLdivergencetothetargetdistri-bution.TheyfullyexploittheapproximationabilityofasetIt...
TransferableClean-LabelPoisoningAttacksonDeepNeuralNetsChenZhu1W.RonnyHuang1AliShafahi1HengduoLi1GavinTaylor2ChristophStuder3TomGoldstein1Abstract2017).Incontrasttoevasionattacks(Biggioetal.,2013;S...
TheEffectofNetworkWidthonStochasticGradientDescentandGeneralization:anEmpiricalStudyDanielS.Park12JaschaSohl-Dickstein1QuocV.Le1SamuelL.Smith3AbstractWilsonetal.,2017;Sagunetal.,2017;Mandtetal.,201...
RiemannianadaptivestochasticgradientalgorithmsonmatrixmanifoldsHiroyukiKasai1PratikJawanpuria2BamdevMishra2AbstractADAM(Kingma&Ba,2015),arguablythemostpopularadaptivegradientmethod,additionallyempl...
RandomWalksonHypergraphswithEdge-DependentVertexWeightsUthsavChitra1BenjaminJRaphael1Abstractbetweengroupsofproteinsinproteincomplexes(Ramadanetal.,2004;Ritzetal.,2014).Hypergraphsareusedinmachinel...
ProbabilityFunctionalDescent:AUnifyingPerspectiveonGANs,VariationalInference,andReinforcementLearningCaseyChu1JoseBlanchet2PeterGlynn2Abstractreinforcementlearning.Thegoalofthispaperistoprovideauni...
PA-GD:ontheConvergenceofPerturbedAlternatingGradientDescenttoSecond-OrderStationaryPointsforStructuredNonconvexOptimizationSongtaoLu1MingyiHong1ZhengdaoWang2AbstractTherearemanywaysofsolvingproblem...
OpenVocabularyLearningonSourceCodewithaGraph–StructuredCacheMilanCvitkovic1BadalSingh2AnimaAnandkumar1AbstractWhilecodecontainsnaturallanguagewordsandphrasesinordertobehuman–readable,codeisnotmea...
onVariationalBoundsofMutualInformationBenPoole1SherjilOzair12Aa¨ronvandenOord3AlexanderA.Alemi1GeorgeTucker1AbstractFigure1.Schematicofvariationalboundsofmutualinformationpresentedinthispaper.Node...