TowardsTightBoundsontheSampleComplexityofAverage-rewardMDPsYujiaJin1AaronSidford1Abstractmakingunderuncertaintyandreinforcementlearning(Puter-man,2014;Sutton&Barto,2018).Itisaprominenttheoret-Wepro...
TightBoundsontheSmallestEigenvalueoftheNeuralTangentKernelforDeepReLUNetworksQuynhNguyen1MarcoMondelli2GuidoMontufar13AbstractWeassumethatthenetworkhasasingleoutput,namelynL=1andWL∈RnL−1.Forconsi...
TightKernelQueryComplexityofKernelRidgeRegressionandKernelk-meansClusteringManuelFerna´ndez1DavidP.Woodruff1TaisukeYasuda2Abstractbetweentwodatapointswiththeirinnerproductafterap-plyingakernelmap,...
SubmodularStreaminginAllItsGlory:TightApproximation,MinimumMemoryandLowAdaptiveComplexityEhsanKazemi1MarkoMitrovic1MortezaZadimoghaddam2SilvioLattanzi2AminKarbasi1Abstractnon-negativesetfunctionf:2...
TightRegretBoundsforBayesianOptimizationinOneDimensionJonathanScarlett1Abstract2010),whoconsiderthecumulativeregret:WeconsidertheproblemofBayesianoptimiza-Ttion(BO)inonedimension,underaGaussianproc...
LyapunovFunctionsforFirst-OrderMethods:TightAutomatedConvergenceGuaranteesAdrienTaylor1BryanVanScoy2LaurentLessard23Abstractmethod,withanenergyofzerocorrespondingtoreachingtheoptimalsolutionof(P).T...
TightBoundsforApproximateCarathéodoryandBeyondVahabMirrokni1RenatoPaesLeme1AdrianVladu2SamChiu-waiWong3Abstractversionstatesthatifoneiswillingtotolerateanerrorof✏in`norm,OD2p/✏2verticessufficeto...