Learn2Hop:LearnedOptimizationonRoughLandscapesWithApplicationstoAtomicStructuralOptimizationAmilMerchant12LukeMetz1SamSchoenholz1EkinDogusCubuk1AbstractFigure1.Schematicdiagramofthedifficultiesofgl...
LossLandscapesofRegularizedLinearAutoencodersDanielKunin1JonathanM.Bloom2AleksandrinaGoeva2CottonSeed2AbstractL(W1,W2)=X−W2W1X2F.Autoencodersareadeeplearningmodelforrepre-ParameterizingLbytheprodu...
AdaGradStepsizes:SharpConvergenceOverNonconvexLandscapesRachelWard12XiaoxiaWu12LéonBottou2AbstractcanbeapproximatedbytheaverageofalargenumberAdaptivegradientmethodssuchasAdaGradandnofcomponentfunc...