PrivateStochasticConvexOptimization:OptimalRatesin1GeometryHilalAsi1VitalyFeldman2TomerKoren3KunalTalwar2AbstractInthisproblem(DP-SCO),givenni.i.d.samplesz1,...,znfromadistributionP,wewishtorelease...
ImprovingLosslessCompressionRatesviaMonteCarloBits-BackCodingYangjunRuan12KarenUllrich23DanielSevero12JamesTownsend4AshishKhisti1ArnaudDoucet5AlirezaMakhzani12ChrisJ.Maddison12Abstract<latexitsha1_...
DistributedSecondOrderMethodswithFastRatesandCompressedCommunicationRustemIslamov12XunQian1PeterRichta´rik1Abstract1.IntroductionWedevelopseveralnewcommunication-efficientTheprevalentparadigmfortr...
OntheGlobalConvergenceRatesofSoftmaxPolicyGradientMethodsJinchengMei12ChenjunXiao1CsabaSzepesva´ri31DaleSchuurmans21Abstracttheyguaranteemonotonicimprovementofthevalue.Asec-ondaryappealisthatpolic...
EvaluatingLossyCompressionRatesofDeepGenerativeModelsSicongHuang123AlirezaMakhzani12YanshuaiCao3RogerGrosse12Abstracttance(FID)(Heuseletal.,2017),whichdonothavenearlythesamedegreeoftheoreticalunder...
ConvergenceRatesofVariationalInferenceinSparseDeepLearningBadr-EddineChérief-Abdellatif1AbstractModernapproximateinferencemainlyreliesonvariationalinference(VI),withsometimesaflavorofsamplingtech-...
SGDwithoutReplacement:SharperRatesforGeneralSmoothConvexFunctionsDheerajNagaraj1PraneethNetrapalli2PrateekJain2Abstractf(x;i):Rd→Risthei-thcomponentfunction.Forex-ample,instandardERManddeeplearnin...
SGD:GeneralAnalysisandImprovedRatesRobertM.Gower1NicolasLoizou2XunQian3AlibekSailanbayev3EgorShulgin4PeterRichta´rik324Abstractwhereeachfi:Rd→Rissmooth(butnotnecessarilyconvex).Further,weassumeth...
RatesofConvergenceforSparseVariationalGaussianProcessRegressionDavidR.Burt1CarlEdwardRasmussen12MarkvanderWilk2AbstractWhilethecomputationalcostofaddinginducingvariablesiswellunderstood,resultsonho...
FastRatesforakNNClassifierRobusttoUnknownAsymmetricLabelNoiseHenryW.J.Reeve1AtaKaba´n1Abstractetal.,2018).InthissettingtheclassicalkNNalgorithmisnolongerconsistent(seeSection5).Mostexistingtheoret...
RatesofConvergenceofSpectralMethodsforGraphonEstimationJiamingXu1Abstractapproximationsofthegraphonfunctionf.Thispaperstudiestheproblemofestimatingthe1.Introductiongraphonfunction–agenerativemecha...
OptimalRatesofSketched-regularizedAlgorithmsforLeast-SquaresRegressionoverHilbertSpacesJunhongLin1VolkanCevher1Abstracttofunctionalregression(Ramsay,2006)andlinearinverseWeinvestigateregularizedalg...
Byzantine-RobustDistributedLearning:TowardsOptimalStatisticalRatesDongYin1YudongChen2KannanRamchandran1PeterBartlett13Abstractasworkermachines(McMahan&Ramage,2017;Konecˇny`etal.,2016).Suchmachines...
Adafactor:AdaptiveLearningRateswithSublinearMemoryCostNoamShazeer1MitchellStern12Abstractvectorsummarizingthehistoryofsquaredgradients,usuallyobtainedthroughsummationasinAdagrad(Duchietal.,Insevera...
UniformConvergenceRatesforKernelDensityEstimationHeinrichJiang1Abstractboundedaswellasdecayassumptionsonthekernelfunc-tions.Moreover,theseboundsholduniformlyoverRdandKerneldensityestimation(KDE)isa...