On the convergence of stochastic gradient descent with adaptive stepsizes

Research output: Chapter in Book/Report/Conference proceedingConference contribution

29 Scopus citations

Abstract

Stochastic gradient descent is the method of choice for large scale optimization of machine learning objective functions. Yet, its performance is greatly variable and heavily depends on the choice of the stepsizes. This has motivated a large body of research on adaptive stepsizes. However, there is currently a gap in our theoretical understanding of these methods, especially in the non-convex setting. In this paper, we start closing this gap: we theoretically analyze in the convex and non-convex settings a generalized version of the AdaGrad stepsizes. We show sufficient conditions for these stepsizes to achieve almost sure asymptotic convergence of the gradients to zero, proving the first guarantee for generalized AdaGrad stepsizes in the nonconvex setting. Moreover, we show that these stepsizes allow to automatically adapt to the level of noise of the stochastic gradients in both the convex and non-convex settings,_ interpolating between O(1/T) and O(1/√T), up to logarithmic terms.
Original languageEnglish (US)
Title of host publicationAISTATS 2019 - 22nd International Conference on Artificial Intelligence and Statistics
PublisherPLMR
StatePublished - Jan 1 2020
Externally publishedYes

Bibliographical note

Generated from Scopus record by KAUST IRTS on 2023-09-25

Fingerprint

Dive into the research topics of 'On the convergence of stochastic gradient descent with adaptive stepsizes'. Together they form a unique fingerprint.

Cite this