RiskBoundsforTransferringRepresentationsWithandWithoutFine-TuningDanielMcNamara1Maria-FlorinaBalcan2Abstractneuralnetworks.1Underthis‘representation-as-a-service’model,ausermayexpecttoaccessthere...
Resource-efficientMachineLearningin2KBRAMfortheInternetofThingsAshishKumar1SaurabhGoyal2ManikVarma1Abstractory.TheBBCMicro:Bithasa32bitARMCortexM0mi-crocontrolleroperatingat16MHzwith16KBSRAMandThis...
RelativeFisherInformationandNaturalGradientforLearningLargeModularModelsKeSun1FrankNielsen23AbstractTheFIMisnotinvariantanddependsontheparameteri-zation.WecanoptionallywriteI(Θ)asIΘ(Θ)toem-Fishe...
RecursivePartitioningforPersonalizationusingObservationalDataNathanKallus1Abstractwhichcanbepotentiallymassiveasinelectronicmedi-calrecords(EMRs)but,atthesametime,lackexperimen-Westudytheproblemofl...
RecoveryGuaranteesforOne-hidden-layerNeuralNetworks∗KaiZhong1ZhaoSong2PrateekJain3PeterL.Bartlett4InderjitS.Dhillon5Abstract1.IntroductionInthispaper,weconsiderregressionproblemsNeuralNetworks(NNs...
RandomFourierFeaturesforKernelRidgeRegression:ApproximationBoundsandStatisticalGuaranteesHaimAvron1MichaelKapralov2CameronMusco3ChristopherMusco3AmeyaVelingker2AmirZandieh2AbstractIntheabove,K2Rn⇥...
RandomFeatureExpansionsforDeepGaussianProcessesKurtCutajar1EdwinV.Bonilla2PietroMichiardi1MaurizioFilippone1Abstract2006)suchthattheoutputofeachlayerofGPsformstheinputtotheGPsatthenextlayer,effecti...
Prox-PDA:TheProximalPrimal-DualAlgorithmforFastDistributedNonconvexOptimizationandLearningOverNetworksMingyiHong1DavoodHajinezhad1Ming-MinZhao2Abstract2016);3)nonconvexutilityfunctionsusedinapplica...
ProtoNN:CompressedandAccuratekNNforResource-scarceDevicesChiragGupta1ArunSaiSuggala12AnkitGoyal13HarshaVardhanSimhadri1BhargaviParanjape1AshishKumar1SaurabhGoyal4RaghavendraUdupa1ManikVarma1Prateek...
ProvableAlternatingGradientDescentforNon-negativeMatrixFactorizationwithStrongCorrelationsYuanzhiLi1YingyuLiang1AbstractBydoingso,onecanavoidcancellationofdifferentfea-turesandimproveinterpretabili...
ProvablyOptimalAlgorithmsforGeneralizedLinearContextualBanditsLihongLi1YuLu2DengyongZhou1Abstractetal.,2009;Lietal.,2010;2012).Intheproblemofper-sonalizednewsrecommendation,thewebsitemustrecom-Cont...
PracticalGauss-NewtonOptimisationforDeepLearningAleksandarBotev1HippolytRitter1DavidBarber12Abstractthegradientsatthecurrentsettingoftheparameterswithvariousheuristicestimatesofthecurvaturefromprev...
PixelCNNModelswithAuxiliaryVariablesforNaturalImageModelingAlexanderKolesnikov1ChristophH.Lampert1Abstracttypemodels(vandenOordetal.,2016a;b;Salimansetal.,2017),haveshowntodeliverthebestperformance...
PartitionedTensorFactorizationsforLearningMixedMembershipModelsZilongTan1SayanMukherjee1Abstractoftheobservationsinatensorform.Inferenceinthisset-tingbecomesaproblemoftensorfactorization.Compu-Wepr...
ParallelandDistributedThompsonSamplingforLarge-scaleAcceleratedExplorationofChemicalSpaceJose´MiguelHerna´ndez-Lobato1JamesRequeima12EdwardO.Pyzer-Knapp34Ala´nAspuru-Guzik3Abstractcompoundsandpo...
OracleComplexityofSecond-OrderMethodsforFinite-SumProblemsYossiArjevani1OhadShamir1AbstractTostudythecomplexityofthisandotheroptimizationproblems,itiscommontoconsideranoraclemodel,Finite-sumoptimiz...
OptimalDensificationforFastandAccurateMinwiseHashingAnshumaliShrivastava1Abstractspeech,andtextarequitepopular(Broder,1997;FetterlyMinwisehashingisafundamentalandoneofetal.,2003).Enrichingfeaturesw...
OptimalAlgorithmsforSmoothandStronglyConvexDistributedOptimizationinNetworksKevinScaman1FrancisBach2Se´bastienBubeck3YinTatLee3LaurentMassoulie´1Abstractnatelocalincrementalimprovementsteps(sucha...
OntheSamplingProblemforKernelQuadratureFranc¸ois-XavierBriol12ChrisJ.Oates34JonCockayne1WilsonYeChen5MarkGirolami24AbstractCarlo(MC)methodscanbeusedtoestimatethenumericalvalueofEqn.1.Aclassicalres...
OnApproximationGuaranteesforGreedyLowRankOptimizationRajivKhanna1EthanR.Elenberg1AlexandrosG.Dimakis1JoydeepGhosh1SahandNegahban2AbstractUnfortunately,theaboveoptimizationcanbecomputation-allytaxin...