5 papers
The Last Iterate of the Projected Stochastic Subgradient Method Pays a Logarithm, Even in Dimension One
Guglielmo Beretta, Tommaso Cesari, Roberto Colomboni
We prove that the projected stochastic subgradient method with the standard anytime stepsizes can suffer expected last-iterate optimization error of order $(\log…
Sharp Dimension Dependence for the Last Iterate of the SubGradient Method
Guglielmo Beretta, Tommaso Cesari, Roberto Colomboni +1
We study the last iterate of the projected subGradient Method (sGM) for convex Lipschitz objectives defined on . We prove that, for a finite horizon and a constan…
New Bounds for the Last Iterate of the Stochastic subGradient Method
Guglielmo Beretta, Tommaso Cesari, Roberto Colomboni +1
We study the last iterate of the stochastic subgradient method for one-dimensional convex Lipschitz objectives. For a fixed horizon , we consider the standard fixed stepsizes $η…
Generalized Pinsker Inequality for Bregman Divergences of Negative Tsallis Entropies
Guglielmo Beretta, Tommaso Cesari, Roberto Colomboni
The Pinsker inequality lower bounds the Kullback--Leibler divergence in terms of total variation and provides a canonical way to convert control…
Vector Flows and the Capacity of a Discrete Memoryless Channel
G. Beretta, G. Chiarot, A. E. Cinà +1
One of the fundamental problems of information theory, since its foundation by Shannon in 1948, has been the computation of the capacity of a discrete memoryless channel, a quantit…