LearninginPOMDPswithMonteCarloTreeSearchSammieKatt1FransA.Oliehoek2ChristopherAmato1Abstracttodecisionmakingbymaintainingaprobabilitydistribu-tionoverpossiblemodelsastheagentactsinanonlinerein-TheP...
ImprovingStochasticPolicyGradientsinContinuousControlwithDeepReinforcementLearningusingtheBetaDistributionPo-WeiChou1DanielMaturana1SebastianScherer1AbstractFigure1.Anexampleofcontinuouscontrolwith...
IdentificationandModelTestinginLinearStructuralEquationModelsusingAuxiliaryVariablesBryantChen1DanielKumor2EliasBareinboim2Abstracteffectisidentifiedrequiresmodelingtheunderlyingcausalstructure,whi...
IdentifytheNashEquilibriuminStaticGameswithRandomPayoffsYichiZhou1JialianLi1JunZhu1Abstracttoestimatethepracticalgamesthroughsimulation.intheempiricalmodeling,pure-strategyprofilesofplayersareWestu...
GradientCoding:AvoidingStragglersinDistributedLearningRashishTandon1QiLei2AlexandrosG.Dimakis3NikosKarampatziakis4AbstractW1W2W3WeproposeanovelcodingtheoreticframeworkD1D2D3formitigatingstragglersi...
GeneralizationandEquilibriuminGenerativeAdversarialNets(GANs)SanjeevArora1RongGe2YingyuLiang1TengyuMa1YiZhang1AbstractFigure1.ProbabilitydensityDrealwithmanypeaksandvalleysItisshownthattrainingofge...
FollowtheMovingLeaderinDeepLearningShuaiZheng1JamesT.Kwok1Abstractever,trainingdeepnetworksisdifficultastheoptimiza-tioncansufferfrompathologicalcurvatureandgetstuckDeepnetworksarehighlynonlinearan...
FairnessinReinforcementLearning⇤ShahinJabbariMatthewJosephMichaelKearnsJamieMorgensternAaronRoth1Abstracttingswherehistoricalcontextcanhaveadistinctinfluenceonthefuture.Forconcreteness,weconsidert...
EfficientRegretMinimizationinNon-ConvexGamesEladHazan1KaranSingh1CyrilZhang1Abstractinthispaperweinvestigatethegeneralizationofthenon-convexstatistical,orbatch,learningmodeltoonlinelearn-Weconsider...
DropoutinferenceinBayesianNeuralNetworkswithAlpha-divergencesYingzhenLi1YarinGal12Abstractinformationcanbeused,forexample,toidentifywhenavi-sionmodelisgivenanadversarialimage(studiedbelow),Toobtain...
DifferentiallyPrivateSubmodularMaximization:DataSummarizationinDisguiseMarkoMitrovic1MarkBun12AndreasKrause3AminKarbasi1AbstractKirchhoff&Bilmes,2014;Siposetal.,2012),crowdteach-ing(Singlaetal.,201...
DifferentiallyPrivateClusteringinHigh-DimensionalEuclideanSpacesMaria-FlorinaBalcan1TravisDick1YingyuLiang2WenlongMou3HongyangZhang1Abstractthequalityofmanydifferentareasofprivatedataanalysis.Westu...
Depth-WidthTradeoffsinApproximatingNaturalFunctionswithNeuralNetworksItaySafran1OhadShamir1Abstractcentempiricalevidencesuggeststhatstandardfeedforwarddeepnetworksarehardertooptimizethanshallowerne...
DecidingHowtoDecide:DynamicRoutinginArtificialNeuralNetworksMasonMcGill1PietroPerona1AbstractMilner,1992),andfacesandotherbehaviorally-relevantstimuliellicitresponsesinanatomicallydistinct,special-...
DARLA:ImprovingZero-ShotTransferinReinforcementLearning111111IrinaHigginsArkaPalAndreiRusuLoicMattheyChristopherBurgessAlexanderPritzel111MatthewBotvinickCharlesBlundellAlexanderLerchnerAbstractef...
ChoiceRank:IdentifyingPreferencesfromNodeTrafficinNetworksLucasMaystre1MatthiasGrossglauser1Abstractitgetsfrompageslinkingtoit).BuildinguponrecentworkbyKumaretal.(2015),wepresentastatisticalframewo...
BeingRobust(inHighDimensions)CanBePracticalIliasDiakonikolas1GautamKamath2DanielM.Kane3JerryLi2AnkurMoitra2AlistairStewart1Abstractgivencomefromanicedistribution,butthatanadversaryhasthepowertoarbi...