(S)GD over Diagonal Linear Networks: Implicit bias, Large Stepsizes and Edge of Stability

Published: 21 Sept 2023, Last Modified: 02 Nov 2023NeurIPS 2023 posterEveryoneRevisionsBibTeX
Keywords: SGD, GD, implicit bias, large stepsizes, edge of stability, diagonal linear networks
TL;DR: We study the convergence and the implicit regularization of (S)GD with large stepsizes over diagonal linear networks.
Abstract: In this paper, we investigate the impact of stochasticity and large stepsizes on the implicit regularisation of gradient descent (GD) and stochastic gradient descent (SGD) over $2$-layer diagonal linear networks. We prove the convergence of GD and SGD with macroscopic stepsizes in an overparametrised regression setting and characterise their solutions through an implicit regularisation problem. Our crisp characterisation leads to qualitative insights about the impact of stochasticity and stepsizes on the recovered solution. Specifically, we show that large stepsizes consistently benefit SGD for sparse regression problems, while they can hinder the recovery of sparse solutions for GD. These effects are magnified for stepsizes in a tight window just below the divergence threshold, in the ``edge of stability'' regime. Our findings are supported by experimental results.
Supplementary Material: pdf
Submission Number: 812