TrainingQuantizedNeuralNetworkstoGlobalOptimalityviaSemidefiniteProgrammingBurakBartan1MertPilanci1Abstractdimensionsusingsemidefiniteprogramming(Bartan&Pi-lanci,2021).Inthiswork,wetakeasimilarconv...
OneforOne,orAllforAll:EquilibriaandOptimalityofCollaborationinFederatedLearningAvrimBlum1NikaHaghtalab2RichardLanasPhillips3HanShao1Abstractpitals(Wenetal.,2019;Powell,2019)anddevices(McMa-han&Rama...
OntheOptimalityofBatchPolicyOptimizationAlgorithmsChenjunXiao12YifanWu3TorLittlemore4BoDai2JinchengMei12LihongLi†5CsabaSzepesvari14DaleSchuurmans12Abstractafixeddatasetofpreviouslycollectedexperie...
GlobalOptimalityBeyondTwoLayers:TrainingDeepReLUNetworksviaConvexProgramsTolgaErgen1MertPilanci1AbstractOutputUnderstandingthefundamentalmechanismbe-Inputhindthesuccessofdeepneuralnetworksisoneofth...
DoublyRobustOff-PolicyActor-Critic:ConvergenceandOptimalityTengyuXu1ZhuoranYang2ZhaoranWang3YingbinLiang1Abstract(Haarnojaetal.,2018),etc.However,thesesuccessesusu-allyrelyontheaccesstoon-policysam...
UniversalAverage-CaseOptimalityofPolyakMomentumDamienScieur1FabianPedregosa2Abstractmizationmethods.Thesimplexmethod,forexample,hasaworst-caseexponentialcomplexity,thatbecomespolyno-Polyakmomentum(...
OntheGlobalOptimalityofModel-AgnosticMeta-Learning:ReinforcementLearningandSupervisedLearningLingxiaoWang1QiCai1ZhuoyanYang2ZhaoranWang1Abstracttimization,wheretheinnerlevelsolveseachsubtaskbasedon...
MedianMatrixCompletion:fromEmbarrassmenttoOptimalityWeidongLiu1XiaojunMao2RaymondK.W.Wong3Abstractflixdatahaslessthan1%ofobservedentriesofamatrixwitharound5×105rowsand2×104customers.Withtech-Inth...
OptimalityImpliesKernelSumClassifiersareStatisticallyEfficientRaphaelA.Meyer1JeanHonorio1AbstractweconsiderbinaryclassificationwithKernelSupportVec-torMachines(SVM),whicharecomputedbysolvingaWeprop...
PartialOptimalityandFastLowerBoundsforWeightedCorrelationClusteringJan-HendrikLange12AndreasKarrenbauer1BjoernAndres134Abstractsolution.Weightedcorrelationclusteringingeneralgraphsishardtosolveexac...