UnsupervisedCo-partSegmentationthroughAssemblyQingzheGao12BinWang2LibinLiu3BaoquanChen3Abstractdifficulttogeneralizetoobjectswithdifferentappearance,lightingorpose,nottomentionunseensubjects.Co-par...
Trainingdata-efficientimagetransformers&distillationthroughattentionHugoTouvron12MatthieuCord12MatthijsDouze1FranciscoMassa1AlexandreSablayrolles1Herve´Je´gou1Abstract⚗↑⚗Recently,neuralnetwork...
OnCharacterizingGANConvergencethroughProximalDualityGapSahilSidheekh1AroofAimen1NarayananC.Krishnan1Abstractmostpopularandsuccessfulofinnovationsforlearningdatadistributions.AGANformulatesthegenera...
ImprovingUltrametricsEmbeddingsthroughCoresetsVincentCohen-Addad1Re´mideJoannisdeVerclos2GuillaumeLagarde3Abstractclusterstructures.Sincethe50s,alargetoolkitofmetricembeddingtechniqueshavebeendeve...
EnhancingRobustnessofNeuralNetworksthroughFourierStabilizationNetanelRaviv1AidanKelley1MichaelGuo1YevgenyVorobeychik1Abstractevasionattacks(Grosseetal.,2016;Li&Vorobeychik,2018;Laskovetal.,2014;Xue...
DifferentiallyPrivateQueryReleasethroughAdaptiveProjectionSergulAydore1WilliamBrown12MichaelKearns13KrishnaramKenthapadi1LucaMelis1AaronRoth13AnkitSiva1AbstractGaussiannoise.Foradatasetofsizen,this...
DescendingthroughaCrowdedValley—BenchmarkingDeepLearningOptimizersRobinM.Schmidt1FrankSchneider1PhilippHennig12Abstract1.IntroductionChoosingtheoptimizerisconsideredtobeamongLarge-scalestochastico...
Ditto:FairandRobustFederatedLearningthroughPersonalizationTianLi1ShengyuanHu1AhmadBeirami2VirginiaSmith1Abstractminimaxoptimizationtofocusontheworst-performingde-vices(Mohrietal.,2019;Huetal.,2020)...
BackdoorScanningforDeepNeuralNetworksthroughK-ArmOptimizationGuangyuShen1YingqiLiu1GuanhongTao1ShengweiAn1QiulingXu1SiyuanCheng1ShiqingMa2XiangyuZhang1Abstractdoor)attackexploitstheuninterpretabili...
ActiveLearningofContinuous-timeBayesianNetworksthroughInterventionsDominikLinzner12HeinzKoeppl13Abstracttolearnthesedependencies.Thisisaproblemwhendataisacquiredunderlimitedresources,whichisthecase...
UnderstandingContrastiveRepresentationLearningthroughAlignmentandUniformityontheHypersphereTongzhouWang1PhillipIsola1AbstractContrastiverepresentationlearninghasbeenout-Alignment:Similarsampleshave...
TrainingBinaryNeuralNetworksthroughLearningwithNoisySupervisionKaiHan12YunheWang2YixingXu2ChunjingXu2EnhuaWu13ChangXu4AbstractFigure1.Frameworkoflearningbinaryneuronswithnoisysuper-vision.Anetworkf...
StructuredPredictionwithPartialLabellingthroughtheInfimumLossVivienCabannes1AlessandroRudi1FrancisBach1AbstractPartiallabellinghasbeenstudiedinthecontextofclassifi-cation(Couretal.,2011;Nguyen&Caru...
ReadyPolicyOne:WorldBuildingthroughActiveLearningPhilipJ.Ball1JackParker-Holder1AldoPacchiano2KrzysztofChoromanski3StephenRoberts1Abstractenvironment)thatcanbeleveragedacrossmanydifferenttasks(tran...
RadioactiveData:TracingthroughTrainingAlexandreSablayrolles12MatthijsDouze1CordeliaSchmid2Herve´Je´gou1Abstract9DQLOODGDWDDatatracingdetermineswhetheraparticularim-7UDLQLQJagedatasethasbeenusedto...
NeuralDatalogthroughTime:InformedTemporalModelingviaLogicalSpecificationHongyuanMei1GuanghuiQin1MinjieXu2JasonEisner1Abstractthedistributionfromwhichthenexteventei+1isdrawn.Alas,whentherelationship...
ImprovingTransformerOptimizationthroughBetterInitializationXiaoShiHuang12FelipePe´rez1JimmyBa32MaksimsVolkovs1Abstractetal.,2019;Sunetal.,2019).Despitethebroadapplications,optimizationintheTransfo...
GuidedLearningofNonconvexModelsthroughSuccessiveFunctionalGradientOptimizationRieJohnson1TongZhang2AbstractOuroriginalmotivationwasfunctionalgradientlearningofadditivemodelsingradientboosting(Fried...
FlexibleandEfficientLong-RangePlanningthroughCuriousExplorationAidanCurtis1MinjianXin2DilipArumugam3KevinFeigelis3DanielYamins3Abstract1.IntroductionIdentifyingalgorithmsthatflexiblyandefficientlyM...
FindingtrainablesparsenetworksthroughNeuralTangentTransferTianlinLiu12FriedemannZenke1Abstract2017).Tothisdate,mostofthestudiesonsparseartificialneuralnetworkshavefocusedonposthocpruning,wherebyDee...