SurrogateLossesforOnlineLearningofStepsizesinStochasticNon-ConvexOptimizationZhenxunZhuang1AshokCutkosky2FrancescoOrabona13Abstractstepsizeηt>0.Inordertoachieveafastconvergence,theStepsizesmustbec...
AdaGradStepsizes:SharpConvergenceOverNonconvexLandscapesRachelWard12XiaoxiaWu12LéonBottou2AbstractcanbeapproximatedbytheaverageofalargenumberAdaptivegradientmethodssuchasAdaGradandnofcomponentfunc...