Zero-ShotTaskGeneralizationwithMulti-TaskDeepReinforcementLearningJunhyukOh1SatinderSingh1HonglakLee12PushmeetKohli3AbstractFigure1:Exampleof3Dworldandinstructions.Theagentistaskedtoexecutelongerse...
VariantsofRMSPropandAdagradwithLogarithmicRegretBoundsMaheshChandraMukkamala12MatthiasHein1AbstractThegoalofthispaperistwofold.First,weproposeSC-AdagradwhichisavariantofAdagradadaptedtotheAdaptiveg...
TheoreticalPropertiesforNeuralNetworkswithWeightMatricesofLowDisplacementRankLiangZhao1SiyuLiao1YanzhiWang2ZheLi2JianTang2BoYuan1AbstractFigure1.ExamplesofcommonlyusedLDR(structured)matri-ces,i.e.,...
TensorDecompositionwithSmoothnessMasaakiImaizumi1KoheiHayashi23AbstractTosolvetheseproblems,thelow-rankassumption,i.e.,giventensorisgeneratedfromasmallnumberoflatentRealdatatensorsaretypicallyhighd...
StrongNP-HardnessforSparseOptimizationwithConcavePenaltyFunctionsYichenChen1DongdongGe2MengdiWang1ZizhuoWang3YinyuYe4HaoYin4AbstractWeareinterestedinthecomputationalcomplexityofProb-lem1undergenera...
Second-OrderKernelOnlineConvexOptimizationwithAdaptiveSketchingDanieleCalandriello1AlessandroLazaric1MichalValko1Abstractminimizetheregret,definedasthedifferencebetweenthelossesofthepredictionsobta...
SchemaNetworks:Zero-shotTransferwithaGenerativeCausalModelofIntuitivePhysicsKenKanskyTomSilverDavidA.Me´lyMohamedEldawyMiguelLa´zaro-GredillaXinghuaLouNimrodDorfmanSzymonSidorScottPhoenixDileepGe...
ScalableGenerativeModelsforMulti-labelLearningwithMissingLabelsVikasJain1NirbhayModhe1PiyushRai1Abstracttisingandrecommendersystems(Prabhu&Varma,2014;Jainetal.,2016),etc.Wepresentascalable,generati...
RobustStructuredEstimationwithSingle-IndexModelsShengChen1ArindamBanerjee1Abstractplicatedscenarios.Tointroducemoreflexibility,oneop-tionistoconsiderthegeneralsingle-indexmodels(SIMs)Inthispaper,we...
RobustProbabilisticModelingwithBayesianDataReweightingYixinWang1AlpKucukelbir1DavidM.Blei1Abstractobservations,oringeneral,measurementsthatdonotbelongDensitytotheprocesswearemodeling.Robustmodelssh...
RobustGaussianGraphicalModelEstimationwithArbitraryCorruptionLingxiaoWang1QuanquanGu1Abstracttheprecisionmatrix⇥⇤correspondstoparameterestima-tion,andspecifyingthenon-zerosetof⇥⇤correspondstoWe...
RiskBoundsforTransferringRepresentationswithandwithoutFine-TuningDanielMcNamara1Maria-FlorinaBalcan2Abstractneuralnetworks.1Underthis‘representation-as-a-service’model,ausermayexpecttoaccessthere...
ReinforcementLearningwithDeepEnergy-BasedPoliciesTuomasHaarnoja1HaoranTang2PieterAbbeel134SergeyLevine1Abstractstochasticpoliciesaredesirableforexploration,thisex-plorationistypicallyattainedheuris...
ProgrammingwithaDifferentiableForthInterpreterMatkoBosˇnjak1TimRockta¨schel2JasonNaradowsky3SebastianRiedel1Abstractpartialproceduralbackgroundknowledge:onemayknowtheroughstructureoftheprogram,or...
PredictionandControlwithTemporalSegmentModelsNikhilMishra1PieterAbbeel12IgorMordatch2Abstracttasksinthesameenvironment.Additionally,learningdif-ferentiabledynamicsmodels(suchasthosebasedonneuralWei...
PixelCNNModelswithAuxiliaryVariablesforNaturalImageModelingAlexanderKolesnikov1ChristophH.Lampert1Abstracttypemodels(vandenOordetal.,2016a;b;Salimansetal.,2017),haveshowntodeliverthebestperformance...
Pain-FreeRandomDifferentialPrivacywithSensitivitySamplingBenjaminI.P.Rubinstein1FrancescoAlda`2Abstracttentrequirementistheneedtoboundglobalsensitivity—aLipschitzconstantofthetarget,non-privatefun...
OnlineLearningwithLocalPermutationsandDelayedFeedbackOhadShamir1LiranSzlak1Abstracttiveandinsomecasesinferiortoalgorithmsnottailoredtocopewithworst-casebehavior.Indeed,anemerginglineofWeproposeanOn...
OnorthogonalityandlearningrecurrentnetworkswithlongtermdependenciesEugeneVorontsov12ChihebTrabelsi12SamuelKadoury13ChrisPal12Abstractstabilizethenormofpropagatingsignalsdirectlybypenal-izingdiffere...
Nystro¨mMethodwithKernelK-means++SamplesasLandmarksDinoOglic12ThomasGa¨rtner2AbstractoreigendecompositionwhichscaleasOn3.ToovercomethiscomputationalshortcomingandscalekernelmethodsWeinvestigate,t...