StochasticSignDescentMethods:NewAlgorithmsandBetterTheoryMherSafaryan1PeterRichtárik12Abstracthencethetrainingdataistypicallysplitandstoredacrossanumberofcomputenodescapableofworkinginparallel.Var...
CRPO:ANewApproachforSafeReinforcementLearningwithConvergenceGuaranteeTengyuXu1YingbinLang1GuanghuiLan2AbstractMind,2019)andrecommendationsystem(Zhengetal.,2018),etc.Inthesesettings,theagentisallowe...
ANewFormalism,MethodandOpenIssuesforZero-ShotCoordinationJohannesTreutlein12MichaelDennis3CasparOesterheld4JakobFoerster125Abstract1.00.91.01.0Inmanycoordinationproblems,independentlyreasoninghuman...
ANewRepresentationofSuccessorFeaturesforTransferacrossDissimilarEnvironmentsMajidAbdolshah1HungLe1ThommenKarimpanalGeorge1SunilGupta1SantuRana1SvethaVenkatesh1Abstractintoindependentsub-domains.How...
VarianceReducedCoordinateDescentwithAcceleration:NewMethodWithaSurprisingApplicationtoFinite-SumProblemsFilipHanzely1DmitryKovalev1PeterRichta´rik1Abstractcontrast,ifψisnotseparable,thecorrespond...
NewOracle-EfficientAlgorithmsforPrivateSyntheticDataReleaseGiuseppeVietri1GraceTian2MarkBun3ThomasSteinke4StevenWu1AbstractmensiondandadatasetD∈XnconsistingofthedataWepresentthreeNewalgorithmsforc...
LEEP:ANewMeasuretoEvaluateTransferabilityofLearnedRepresentationsCuongV.Nguyen1TalHassner2MatthiasSeeger1CedricArchambeau1Abstractchoosegoodsourcemodelsforagiventargettask(Achilleetal.,2019;Baoetal...
GeneralizationtoNewActionsinReinforcementLearningAyushJain1AndrewSzot1JosephJ.Lim1AbstractActionAfundamentaltraitofintelligenceistheabil-GoalGoalitytoachievegoalsinthefaceofnovelcircum-stances,such...
FastAdaptationtoNewEnvironmentsviaPolicy-DynamicsValueFunctionsRobertaRaileanu1MaxGoldstein1ArthurSzlam2RobFergus1Abstractmighthavetoadjustitsbehaviordependingonweathercon-ditions,oraprostheticcont...
Divide,Conquer,andCombine:aNewInferenceStrategyforProbabilisticProgramswithStochasticSupportYuanZhou1HongseokYang2YeeWhyeTeh3TomRainforth3AbstractFigure1.MAPestimateofthemeansandcovariancesofaGaus-...
ANewregretanalysisforAdam-typealgorithmsAhmetAlacaoglu1YuraMalitsky1PanayotisMertikopoulos23VolkanCevher1AbstractOnecanwonderwhetherthereisaninherentobstacle–intheproposedmethodsorthesetting–whic...
RevisitingtheSoftmaxBellmanOperator:NewBenefitsandNewPerspectiveZhaoSong1RonaldE.Parr1LawrenceCarin1Abstracttivatestheuseofexploratoryandpotentiallysub-optimalactionsduringlearning,andonecommonly-u...
NewResultsonInformationTheoreticClusteringFerdinandoCicalese1EduardoLaber2LucasMurtinho2Abstracttropy)thatestimatethedissimilarityofagroupofitems(see,e.g.,(Dhillonetal.,2003)andreferencestherein)In...
DynamicLearningwithFrequentNewProductLaunches:ASequentialMultinomialLogitBanditProblemJunyuCao1WeiSun2Abstractmarketdynamicssoastoimprovelonger-termprofitability,yettheymayhavetosacrificeshort-term...
FittingNewSpeakersBasedonaShortUntranscribedSampleEliyaNachmani12AdamPolyak1YanivTaigman1LiorWolf12AbstractInthiswork,weproposeaTTSnetworkthatisdesignedtofitaNewvoicebasedonalimitedamountofdataandL...
UncorrelationandEvenness:aNewDiversity-PromotingRegularizerPengtaoXie12AartiSingh1EricP.Xing2AbstractFirst,undermanycircumstances,thefrequencyofpatternsishighlyimbalanced.Somepatternshaveveryhighfr...
InnovationPursuit:ANewApproachtotheSubspaceClusteringProblemMostafaRahmani1GeorgeAtia1Abstracttothesesubspaces.Subspaceclusteringnaturallyarisesinmanymachinelearninganddataanalysisproblems,in-Thisp...