Leonardo Galli, Daniela Angela Parletta, Andrea Paudice
原文摘要
In this paper, we study stochastic gradient descent in the smooth interpolation regime, where the population objective shares a common minimizer with the sample losses. In contrast to prior work, which has largely focused on in-expectation analysis, we obtain high-probability convergence guarantees under the same assumptions. In the convex case, we prove a fast rate of order $\log(β^{-1}) \cdot n^{-1}$ after $n$ iterations, with failure probability at most $β$. In the strongly convex case, we derive a family of moment-based bounds of order $β^{-p} \cdot \exp(-\mathcal{O}(p \cdot n))$ for $p \in (0,1]$, and complement them with a matching lower bound showing that the corresponding deterioration in the exponential rate is essentially unavoidable. In both settings, our theoretical results are unified across a broad class of step-size schedules. Building on this analysis, we extend our results to a low-noise regime and to nonconvex objectives satisfying the Polyak-Lojasiewicz condition.
We study the almost sure convergence of the last-iterate optimization error of Stochastic (sub)Gradient Descent (SGD) for convex objectives. For smooth objectives, we establish convergence rates of order $k^{-1/3}$ when a stationary minimizer exists, and $k^{-1/4}$ otherwise, up to polylogarithmic factors, where $k$ denotes the iteration counter. We further show that the latter rate extends to nonsmooth convex objectives. All these rates are achieved by deterministic, anytime stepsize policies. Finally, we establish matching lower bounds showing that the exponents $-1/3$ and $-1/4$ are optimal, up to polylogarithmic factors, for the corresponding problem classes.
Adversarial Training (AT) is an important method for improving model robustness. We study the dynamics of deterministic Gradient Descent (GD) in AT. Most existing works in this direction focus on the small or vanishing step sizes regime, where GD resembles a continuous-time flow. However, such step sizes can make training inefficient and are rarely used in practice. In this paper, we study the AT dynamics of GD with large constant step sizes. We first study GD in adversarial linear logistic regression, focusing on two aspects: where the model parameters converge (implicit bias) and how quickly the losses decrease (optimization). We then turn to neural networks training and study how the sharpness of losses evolve during AT. Our results advance the understanding of the behaviors of GD with large step sizes in AT and highlight three key differences from its behavior in Standard Training (ST): (i) the implicit bias gap between AT and ST, previously established only for small step sizes, persists for arbitrarily large constant step sizes, (ii) different from ST, there is no uniform rate of convergence for the optimization error in AT, and (iii) sharpness can exhibit two-stage dynamics that differ from the Edge-of-Stability behavior observed in ST.