Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Li, Chris Junchi, Mou, Wenlong, Wainwright, Martin J., Jordan, Michael I.
Format: Preprint
Veröffentlicht: 2020
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2008.12690
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912032044875776
author Li, Chris Junchi
Mou, Wenlong
Wainwright, Martin J.
Jordan, Michael I.
author_facet Li, Chris Junchi
Mou, Wenlong
Wainwright, Martin J.
Jordan, Michael I.
contents We study the problem of solving strongly convex and smooth unconstrained optimization problems using stochastic first-order algorithms. We devise a novel algorithm, referred to as Recursive One-Over-T SGD (ROOT-SGD), based on an easily implementable, recursive averaging of past stochastic gradients. We prove that it simultaneously achieves state-of-the-art performance in both a finite-sample, nonasymptotic sense and an asymptotic sense. On the non-asymptotic side, we prove risk bounds on the last iterate of ROOT-SGD with leading-order terms that match the optimal statistical risk with a unity pre-factor, along with a higher-order term that scales at the sharp rate of $O(n^{-3/2})$ under the Lipschitz condition on the Hessian matrix. On the asymptotic side, we show that when a mild, one-point Hessian continuity condition is imposed, the rescaled last iterate of (multi-epoch) ROOT-SGD converges asymptotically to a Gaussian limit with the Cramér-Rao optimal asymptotic covariance, for a broad range of step-size choices.
format Preprint
id arxiv_https___arxiv_org_abs_2008_12690
institution arXiv
publishDate 2020
record_format arxiv
spellingShingle ROOT-SGD: Sharp Nonasymptotics and Near-Optimal Asymptotics in a Single Algorithm
Li, Chris Junchi
Mou, Wenlong
Wainwright, Martin J.
Jordan, Michael I.
Optimization and Control
Machine Learning
We study the problem of solving strongly convex and smooth unconstrained optimization problems using stochastic first-order algorithms. We devise a novel algorithm, referred to as Recursive One-Over-T SGD (ROOT-SGD), based on an easily implementable, recursive averaging of past stochastic gradients. We prove that it simultaneously achieves state-of-the-art performance in both a finite-sample, nonasymptotic sense and an asymptotic sense. On the non-asymptotic side, we prove risk bounds on the last iterate of ROOT-SGD with leading-order terms that match the optimal statistical risk with a unity pre-factor, along with a higher-order term that scales at the sharp rate of $O(n^{-3/2})$ under the Lipschitz condition on the Hessian matrix. On the asymptotic side, we show that when a mild, one-point Hessian continuity condition is imposed, the rescaled last iterate of (multi-epoch) ROOT-SGD converges asymptotically to a Gaussian limit with the Cramér-Rao optimal asymptotic covariance, for a broad range of step-size choices.
title ROOT-SGD: Sharp Nonasymptotics and Near-Optimal Asymptotics in a Single Algorithm
topic Optimization and Control
Machine Learning
url https://arxiv.org/abs/2008.12690