Revisiting Gradient Clipping: Stochastic bias and tight convergence guarantees - Inria EPFL
Conference Papers Year : 2023

Revisiting Gradient Clipping: Stochastic bias and tight convergence guarantees

Abstract

Gradient clipping is a popular modification to standard (stochastic) gradient descent, at every iteration limiting the gradient norm to a certain value $c >0$. It is widely used for example for stabilizing the training of deep learning models (Goodfellow et al., 2016), or for enforcing differential privacy (Abadi et al., 2016). Despite popularity and simplicity of the clipping mechanism, its convergence guarantees often require specific values of $c$ and strong noise assumptions. In this paper, we give convergence guarantees that show precise dependence on arbitrary clipping thresholds $c$ and show that our guarantees are tight with both deterministic and stochastic gradients. In particular, we show that (i) for deterministic gradient descent, the clipping threshold only affects the higher-order terms of convergence, (ii) in the stochastic setting convergence to the true optimum cannot be guaranteed under the standard noise assumption, even under arbitrary small step-sizes. We give matching upper and lower bounds for convergence of the gradient norm when running clipped SGD, and illustrate these results with experiments.
Fichier principal
Vignette du fichier
2305.01588.pdf (784.06 Ko) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-04107297 , version 1 (26-05-2023)

Licence

Identifiers

Cite

Anastasia Koloskova, Hadrien Hendrikx, Sebastian U. Stich. Revisiting Gradient Clipping: Stochastic bias and tight convergence guarantees. ICML 2023 - 40th International Conference on Machine Learning, Jul 2023, Honolulu, Hawaii,, United States. pp.1-19. ⟨hal-04107297⟩
153 View
154 Download

Altmetric

Share

More