Understanding Gradient Descent on Edge of Stability in Deep Learning

• URL: http://arxiv.org/abs/2205.09745v1
• Date: Thu, 19 May 2022 17:57:01 GMT
• Title: Understanding Gradient Descent on Edge of Stability in Deep Learning
• Authors: Sanjeev Arora, Zhiyuan Li, Abhishek Panigrahi
• Abstract summary: This paper mathematically analyzes a new mechanism of implicit regularization in the EoS phase, whereby GD updates due to non-smooth loss landscape turn out to evolve along some deterministic flow on the manifold of minimum loss. The above theoretical results have been corroborated by an experimental study.
• Score: 32.03036040349019
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Deep learning experiments in Cohen et al. (2021) using deterministic Gradient Descent (GD) revealed an {\em Edge of Stability (EoS)} phase when learning rate (LR) and sharpness (\emph{i.e.}, the largest eigenvalue of Hessian) no longer behave as in traditional optimization. Sharpness stabilizes around $2/$LR and loss goes up and down across iterations, yet still with an overall downward trend. The current paper mathematically analyzes a new mechanism of implicit regularization in the EoS phase, whereby GD updates due to non-smooth loss landscape turn out to evolve along some deterministic flow on the manifold of minimum loss. This is in contrast to many previous results about implicit bias either relying on infinitesimal updates or noise in gradient. Formally, for any smooth function $L$ with certain regularity condition, this effect is demonstrated for (1) {\em Normalized GD}, i.e., GD with a varying LR $ \eta_t =\frac{ \eta }{ || \nabla L(x(t)) || } $ and loss $L$; (2) GD with constant LR and loss $\sqrt{L}$. Both provably enter the Edge of Stability, with the associated flow on the manifold minimizing $\lambda_{\max}(\nabla^2 L)$. The above theoretical results have been corroborated by an experimental study.

Related papers

• Methods for Convex $(L_0,L_1)$-Smooth Optimization: Clipping, Acceleration, and Adaptivity [50.25258834153574]
  We focus on the class of (strongly) convex $(L0)$-smooth functions and derive new convergence guarantees for several existing methods. In particular, we derive improved convergence rates for Gradient Descent with smoothnessed Gradient Clipping and for Gradient Descent with Polyak Stepsizes.
  arXiv  Detail & Related papers  (2024-09-23T13:11:37Z)
• Large Stepsize Gradient Descent for Logistic Loss: Non-Monotonicity of the Loss Improves Optimization Efficiency [47.8739414267201]
  We consider gradient descent (GD) with a constant stepsize applied to logistic regression with linearly separable data. We show that GD exits this initial oscillatory phase rapidly -- in $mathcalO(eta)$ steps -- and subsequently achieves an $tildemathcalO (1 / (eta t) )$ convergence rate. Our results imply that, given a budget of $T$ steps, GD can achieve an accelerated loss of $tildemathcalO (1/T2)$ with an aggressive stepsize
  arXiv  Detail & Related papers  (2024-02-24T23:10:28Z)
• Nearly Minimax Optimal Regret for Learning Linear Mixture Stochastic Shortest Path [80.60592344361073]
  We study the Shortest Path (SSP) problem with a linear mixture transition kernel. An agent repeatedly interacts with a environment and seeks to reach certain goal state while minimizing the cumulative cost. Existing works often assume a strictly positive lower bound of the iteration cost function or an upper bound of the expected length for the optimal policy.
  arXiv  Detail & Related papers  (2024-02-14T07:52:00Z)
• Lower Generalization Bounds for GD and SGD in Smooth Stochastic Convex Optimization [9.019243171993553]
  Training steps $T$ and step-size $eta$ might affect certify in smooth convex optimization (SCO) problems. We first provide tight excess risk lower bounds for Gradient Descent (GD) and Gradient Descent (SGD) Recent works show better rates can be attained but the improvement is reduced when training time is long.
  arXiv  Detail & Related papers  (2023-03-19T20:24:33Z)
• Variance-reduced Clipping for Non-convex Optimization [24.765794811146144]
  Gradient clipping is a technique used in deep learning applications such as large-scale language modeling. Recent experimental training have a fairly special behavior in that it mitigates order complexity.
  arXiv  Detail & Related papers  (2023-03-02T00:57:38Z)
• Generalization Bounds for Gradient Methods via Discrete and Continuous Prior [8.76346911214414]
  We show a new high probability generalization bound of order $O(frac1n + fracL2n2sum_t=1T(gamma_t/varepsilon_t)2)$ for gradient Langevin Dynamics (GLD) We can also obtain new bounds for certain variants of SGD.
  arXiv  Detail & Related papers  (2022-05-27T07:23:01Z)
• Improved Convergence Rate of Stochastic Gradient Langevin Dynamics with Variance Reduction and its Application to Optimization [50.83356836818667]
  gradient Langevin Dynamics is one of the most fundamental algorithms to solve non-eps optimization problems. In this paper, we show two variants of this kind, namely the Variance Reduced Langevin Dynamics and the Recursive Gradient Langevin Dynamics.
  arXiv  Detail & Related papers  (2022-03-30T11:39:00Z)
• Black-Box Generalization [31.80268332522017]
  We provide the first error analysis for black-box learning through derivative generalization. We show both generalization are independent $d$, $K$ and under appropriate choices a slightly decreased learning rate.
  arXiv  Detail & Related papers  (2022-02-14T17:14:48Z)
• What Happens after SGD Reaches Zero Loss? --A Mathematical Framework [35.31946061894308]
  Understanding the implicit bias of Gradient Descent (SGD) is one of the key challenges in deep learning. This paper gives a general framework for such analysis by adapting ideas from Katzenberger (1991). It yields some new results: (1) a global analysis of the implicit bias valid for $eta-2$ steps, in contrast to the local analysis of Blanc et al. ( 2020) that is only valid for $eta-1.6$ steps and (2) allowing arbitrary noise covariance.
  arXiv  Detail & Related papers  (2021-10-13T17:50:46Z)
• Direction Matters: On the Implicit Bias of Stochastic Gradient Descent with Moderate Learning Rate [105.62979485062756]
  This paper attempts to characterize the particular regularization effect of SGD in the moderate learning rate regime. We show that SGD converges along the large eigenvalue directions of the data matrix, while GD goes after the small eigenvalue directions.
  arXiv  Detail & Related papers  (2020-11-04T21:07:52Z)
• Optimal Robust Linear Regression in Nearly Linear Time [97.11565882347772]
  We study the problem of high-dimensional robust linear regression where a learner is given access to $n$ samples from the generative model $Y = langle X,w* rangle + epsilon$ We propose estimators for this problem under two settings: (i) $X$ is L4-L2 hypercontractive, $mathbbE [XXtop]$ has bounded condition number and $epsilon$ has bounded variance and (ii) $X$ is sub-Gaussian with identity second moment and $epsilon$ is
  arXiv  Detail & Related papers  (2020-07-16T06:44:44Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.