Circuit-BasedIntrinsicMethodstoDetectOverfittingSatrajitChatterjee1AlanMishchenko2Abstractknowledge,suchas,theperformanceofthemodelonexam-plesheldoutfromthetrainingprocess,detailsoftheprocessThefoc...
Black-boxMethodsforRestoringMonotonicityEvangeliaGergatsouli1BrendanLucier2ChristosTzamos1Abstractoftheknapsackincreases,thenewsetofitemsabouttobeselectedareexpectedtobeatleastasvaluableasbefore.In...
AndersonAccelerationofProximalGradientMethodsVienV.Mai1MikaelJohansson1Abstractrameters;slightlyover-orunder-estimatingthestrongcon-vexityconstantcanhaveasevereeffectontheoverallper-Andersonacceler...
AcceleratedStochasticGradient-freeandProjection-freeMethodsFeihuHuang12LueTao12SongcanChen12Abstract1.IntroductionInthepaper,weproposeaclassofacceleratedInthepaper,wefocusonsolvingthefollowingconst...
UnifyingOrthogonalMonteCarloMethodsKrzysztofChoromanski1MarkRowland2WenyuChen3AdrianWeller24Abstractetal.,2018b).WerefertotheclassofMethodsusingsuchorthogonalcouplingsasorthogonalMonteCarlo(OMC).Ma...
SimpleStochasticGradientMethodsforNon-SmoothNon-ConvexRegularizedOptimizationMichaelR.Metel1AkikoTakeda12Abstractwherefj(w)=F(w,ξj)andhasaLipschitzcontinuousgradient.Ourworkfocusesonstochasticgrad...
Plug-and-PlayMethodsProvablyConvergewithProperlyTrainedDenoisersErnestK.Ryu1JialinLiu1SichengWang2XiaohanChen2ZhangyangWang2WotaoYin1Abstractmeasurementsoftheimage,isencodedinf(x).Sof(x)issmallifxa...
MakingDeepQ-learningMethodsRobusttoTimeDiscretizationCorentinTallec1Le´onardBlier12YannOllivier2Abstractpreventstransferfromimperfectsimulatorstorealworldscenarios.Despiteremarkablesuccesses,DeepR...
EscapingSaddlePointswithAdaptiveGradientMethodsMatthewStaib12SashankReddi3SatyenKale3SanjivKumar3SuvritSra1AbstractAdagradupdatestheparametersinthefollowingmanner:AdaptiveMethodssuchasAdamandRMSPro...
ConditionalGradientMethodsviaStochasticPath-IntegratedDifferentialEstimatorAlpYurtsever1SuvritSra2VolkanCevher1AbstractarelatedworkbyHazan&Luo(2016)),constraineddeeplearningproblems(e.g.,Ravietal.(...
AcceleratedLinearConvergenceofStochasticMomentumMethodsinWassersteinDistancesBugraCan1MertGurbuzbalaban1LingjiongZhu2Abstractsupervisedlearningincludelinearandnon-linearregressionproblems,supportve...
StochasticVariance-ReducedHamiltonMonteCarloMethodsDifanZou1PanXu1QuanquanGu1Abstractconvergestoitsstationarydistribution,a.k.a.,theGibbsmeasure⇡/exp(f(x)).Notethat⇡issmoothandWeproposeafaststoch...
StochasticVariance-ReducedCubicRegularizedNewtonMethodsDongruoZhou1PanXu1QuanquanGu1Abstracttheirextensions(Ghadimi&Lan,2013;Reddietal.,2016a;Allen-Zhu&Hazan,2016;Ghadimi&Lan,2016).Never-Weproposea...
SADAGRAD:StronglyAdaptiveStochasticGradientMethodsZaiyiChen12YiXu2EnhongChen1TianbaoYang2Abstractiterations.Ithasreceivedtremendousinterestsforsolvingbigdatalearningproblems(e.g.,see(Deanetal.,2012...
RatesofConvergenceofSpectralMethodsforGraphonEstimationJiamingXu1Abstractapproximationsofthegraphonfunctionf.Thispaperstudiestheproblemofestimatingthe1.Introductiongraphonfunction–agenerativemecha...
LyapunovFunctionsforFirst-OrderMethods:TightAutomatedConvergenceGuaranteesAdrienTaylor1BryanVanScoy2LaurentLessard23Abstractmethod,withanenergyofzerocorrespondingtoreachingtheoptimalsolutionof(P).T...
Level-SetMethodsforFinite-SumConstrainedConvexOptimizationQihangLin1RunchaoMa1TianbaoYang2AbstractAsolutionx¯∈Xisε-optimaliff0(x¯)−f∗≤εandε-feasibleifmaxi=1,...,m[fi(x¯)−ri]≤ε.Weconsid...
GlobalConvergenceofPolicyGradientMethodsfortheLinearQuadraticRegulatorMaryamFazel1RongGe2ShamM.Kakade1MehranMesbahi1Abstract2016)andAtarigameplaying(Mnihetal.,2015).Deepreinforcementlearning(DeepRL...
FastGradient-BasedMethodswithExponentialRate:AHybridControlFrameworkArmanSharifiKolarijani1PeymanMohajerinEsfahani1Tama´sKeciczky1Abstracttheanalysisanddesignofoptimizationalgorithms.Con-sideranit...
DynamicRegretofStronglyAdaptiveMethodsLijunZhang1TianbaoYang2RongJin3Zhi-HuaZhou1Abstractincurredbythelearnerandthatofthebestfixeddecisioninhindsight,i.e.,Tocopewithchangingenvironments,recentde-ve...