UnderstandingPriorsinBayesianNeuralNetworksattheUnitLevelMariiaVladimirova12JakobVerbeek1PabloMesejo3JulyanArbel1AbstractworksonBayesianneuralnetworks(Neal,1992;MacKay,1992).Thereisalargevarietyofa...
UnderstandingandcorrectingpathologiesinthetrainingoflearnedoptimizersLukeMetz1NiruMaheswaranathan1JeremyNixon1C.DanielFreeman1JaschaSohl-Dickstein1Abstractperparametersearch.Ontheotherhand,acomplem...
UnderstandingandControllingMemoryinRecurrentNeuralNetworksDoronHaviv12AlexnaderRivkind234OmriBarak23Abstract1.introductionTobeeffectiveinsequentialdataprocessing,Re-RecurrentNeuralNetworks(RNN)aret...
TransferofSamplesinPolicySearchviaMultipleImportanceSamplingAndreaTirinzoni1MattiaSalvini1MarcelloRestelli1Abstractagentissupposedtoreuseknowledgeacquiredfromasetofsourcetaskstoacceleratethelearnin...
TraditionalandHeavyTailedSelfRegularizationinNeuralNetworkModelsCharlesH.Martin1MichaelW.Mahoney2Abstracttheorydidnotapplytothesesystems(Vapniketal.,1994).Itwasoriginallyassumedthatlocalminimainthe...
TowardsAccurateModelSelectioninDeepUnsupervisedDomainAdaptationKaichaoYou12XimeiWang12MingshengLong12MichaelI.Jordan3Abstractnaturalinrecognitiontaskswheredeepmodelshaveshowntheirsuperiority(Longet...
TowardsUnderstandingtheImportanceofNoiseinTrainingNeuralNetworksMoZhou1TianyiLiu2YanLi2DachaoLin1EnluZhou2TuoZhao2AbstractSimplefirstorderalgorithmssuchasStochasticGradientDescent(SGD)anditsvariant...
TowardControllingDiscriminationinOnlineAdAuctionsL.ElisaCelis1AnayMehrotra2NisheethK.Vishnoi3Abstracttheadvertiserischarged(Muthukrishnan,2009;Yuanetal.,2012;Varian,2007).Asitisnotpracticalforadver...
TighterProblem-DependentRegretBoundsinReinforcementLearningwithoutDomainKnowledgeusingValueFunctionBoundsAndreaZanette1EmmaBrunskill2AbstractFortunatelyinpracticereinforcementlearningalgorithmsof-t...
TheValueFunctionPolytopeinReinforcementLearningRobertDadashi1AdrienAliTa¨ıga12NicolasLeRoux1DaleSchuurmans13MarcG.Bellemare1AbstractLinetheorem.Weshowthatpoliciesthatagreeonallbutonestategenerate...
Theinformation-TheoreticValueofUnlabeledDatainSemi-SupervisedLearningAlexanderGolovnev1Da´vidPa´l2Bala´zsSzo¨re´nyi2Abstractofalgorithmsindexedbythe(uncountablymany)distri-butionsoverthedomain...
TheAnisotropicNoiseinStochasticGradientDescent:ItsBehaviorofEscapingfromSharpMinimaandRegularizationEffectsZhanxingZhu123JingfengWu1BingYu1LeiWu1JinwenMa1Abstract90Understandingthebehaviorofstochas...
SWALP:StochasticWeightAveraginginLow-PrecisionTrainingGuandaoYang1TianyiZhang1PolinaKirichenko1JunwenBai1AndrewGordonWilson1ChristopherDeSa1Abstractandaccumulategradientinformationinhigherprecision...
SubmodularStreaminginAllItsGlory:TightApproximation,MinimumMemoryandLowAdaptiveComplexityEhsanKazemi1MarkoMitrovic1MortezaZadimoghaddam2SilvioLattanzi2AminKarbasi1Abstractnon-negativesetfunctionf:2...
StatisticsandSamplesinDistributionalReinforcementLearningMarkRowland1RobertDadashi2SaurabhKumar2Re´miMunos1MarcG.Bellemare2WillDabney1AbstractthatDRLalgorithmscanbeviewedascombiningastatisti-cales...
StaticAutomaticBatchinginTensorFlowAshishAgarwal1Abstractterpretedlanguageslikepythoncanmakethisbottleneckworse.Dynamicneuralnetworksarebecomingincreas-inglycommon,andyetitishardtoimplementGiventhe...
Self-SimilarEpochs:ValueinArrangementEliavBuchnik12EdithCohen21AvinatanHassidim2YossiMatias2Abstractbroad:entitiescanbeofoneormultipletypesandexampleassociationsusedfortrainingcanberaworpreprocesse...
ScalableLearninginReproducingKernelKre˘ınSpacesDinoOglic1ThomasGärtner2AbstractandSmola,2001).TheNyströmmethod(Nyström,1930;SmolaandSchölkopf,2000;WilliamsandSeeger,2001)isWeprovidethefirstma...
RegularizationinDirectableEnvironmentswithApplicationtoTetrisJanMalteLichtenberg1O¨zgu¨rS¸ims¸ek1Abstractonregularization.Specifically,weproposeamodelthatintroducesabiastowardgivingallfeaturese...
RehashingKernelEvaluationinHighDimensionsParisSiminelakis1KexinRong1PeterBailis1MosesCharikar1PhilipLevis1Abstract(a)kernel(b)difficultcase(c)simplecaseKernelmethodsareeffectivebutdonotscalewellFig...