ActionRobustReinforcementLearningandApplicationsinContinuousControlChenTessler1YonathanEfroni1ShieMannor1AbstractTheadvantageofrobustpoliciesishighlightedwhencon-sideringimperfectmodels,acommonscen...
ATail-indexAnalysisofStochasticGradientNoiseinDeepNeuralNetworksUmutS¸ims¸ekli1LeventSagun2MertGu¨rbu¨zbalaban3Abstractmanyapplicationdomains(LeCunetal.,2015;Krizhevskyetal.,2012;Hintonetal.,20...
AStatisticalinvestigationofLongMemoryinLanguageandMusicAlexanderGreaves-Tunnell1ZaidHarchaoui1Abstractundoubtedlyhelpful,suchheuristicsarerarelydefinedwithrespecttoanunderlyingmathematicalorstatist...
ALarge-ScaleStudyonRegularizationandNormalizationinGANsKarolKurach1MarioLucic1XiaohuaZhai1MarcinMichalski1SylvainGelly1Abstractfromthetruedistributionorweresynthesizedbythegenera-tor.Thesolutiontot...
AFrameworkforBayesianOptimizationinEmbeddedSubspacesAlexanderMunteanu1AminNayebi2MatthiasPoloczek32Abstracttheobjectivefunction.Thus,itisnotsurprisingthatex-pandingBOtohigher-dimensionalsearchspace...
Zero-ShotKnowledgeDistillationinDeepNetworksGauravKumarNayak1KondaReddyMopuri2VaisakhShaj3R.VenkateshBabu1AnirbanChakraborty1Abstractlentperformance,buttheycanbehugeandcomputationallyexpensive.Henc...
WidthProvablyMattersinOptimizationforDeepLinearNeuralNetworksSimonS.Du1WeiHu2Abstractconvergestoglobalminimumunderfurtherassumptionsonbothdataandglobalminimum.TheseresultsrequireWeprovethatforanL-l...
WeakDetectionofSignalintheSpikedWignerModelHyeWonChung1JiOonLee2Abstractwherethesignalx∈RNandHisanN×NWignerma-trix.(SeeDefinitions1and2.)ThespikedWignermodelWeconsidertheproblemofdetectingthepres...
ABaselineforAnyOrderGradientEstimationinStochasticComputationGraphsJingkaiMao1JakobFoerster2TimRockta¨schel3MaruanAl-Shedivat4GregoryFarquhar2ShimonWhiteson2Abstract1.introductionByenablingcorrect...
TransferinDeepReinforcementLearningUsingSuccessorFeaturesandGeneralisedPolicyImprovementAndre´Barreto1DianaBorsa1JohnQuan1TomSchaul1DavidSilver1MatteoHessel1DanielMankowitz1AugustinZˇ´ıdek1Re´...
TimeLimitsinReinforcementLearningFabioPardo1ArashTavakoli1VitalyLevdik1PetarKormushev1Abstractintheenvironmentwhichinturnprovidesarepresenta-tionSt+1ofthesuccessorstateandarewardsignalRt+1.inreinfo...
TightRegretBoundsforBayesianOptimizationinOneDimensionJonathanScarlett1Abstract2010),whoconsiderthecumulativeregret:WeconsidertheproblemofBayesianoptimiza-Ttion(BO)inonedimension,underaGaussianproc...
TheMirageofAction-DependentBaselinesinReinforcementLearningGeorgeTucker1SuryaBhupatiraju12ShixiangGu134RichardE.Turner3ZoubinGhahramani35SergeyLevine16Abstractetal.,2015a;2017)areaclassofmodel-free...
TheHiddenVulnerabilityofDistributedLearninginByzantiumElMahdiElMhamdi1RachidGuerraoui1Se´bastienRouault1AbstractQ,dependingonaparameterx,ifonekeepsupdatingxintheoppositedirectionofthegradientofQ,w...
TheEdgeDensityBarrier:Computational-StatisticalTradeoffsinCombinatorialinferenceHaoLu1YuanCao1ZhuoranYang1JunweiLu2HanLiu3ZhaoranWang4Abstractinbioinformatics(Friedman,2004),informationretrieval(We...
StreamingPrincipalComponentAnalysisinNoisySettingsTeodorV.Marinov1PooryaMianjy1RamanArora1AbstractBigdataischaracterizednotonlybyitssheer“volume”butalsobyits“veracity”,orthelackthereof.Mostbigd...
SpuriousLocalMinimaareCommoninTwo-LayerReLUNeuralNetworksItaySafran1OhadShamir1Abstractlearning,andtensordecomposition,donothavespuriouslocalminimaundersuitableassumptions,inwhichcaselo-Weconsidert...
ResidualUnfairnessinFairMachineLearningfromPrejudicedDataNathanKallus1AngelaZhou2Abstractnewquestionsaboutthepossibleharmsoflearningfromdatawhichissubjecttohistoricalbias.Unlikeclean-cutpre-Recentw...
RevealingCommonStatisticalBehaviorsinHeterogeneousPopulationsAndreyZhitnikov1RotemMulayoff1TomerMichaeli1Abstractanalysis(Hagleretal.,2006),etc.Groupanalysesoftenrelyontheassumptionthatallsubjectsi...
QuantTree:HistogramsforChangeDetectioninMultivariateDataStreamsGiacomoBoracchi1DiegoCarrera1CristianoCervellera2DaniloMaccio`2Abstractmostchange-detectiontestsintheliterature(Basseville&Nikiforov,1...