Nonlinear Acceleration of Deep Neural Networks

Damien Scieur 1, 2 Edouard Oyallon 3, 4 Alexandre D 'Aspremont 1, 2 Francis Bach 1, 2
1 SIERRA - Statistical Machine Learning and Parsimony
DI-ENS - Département d'informatique de l'École normale supérieure, CNRS - Centre National de la Recherche Scientifique, Inria de Paris
Abstract : Regularized nonlinear acceleration (RNA) is a generic extrapolation scheme for optimization methods, with marginal computational overhead. It aims to improve convergence using only the iterates of simple iterative algorithms. However, so far its application to optimization was theoretically limited to gradient descent and other single-step algorithms. Here, we adapt RNA to a much broader setting including stochastic gradient with momentum and Nesterov's fast gradient. We use it to train deep neural networks, and empirically observe that extrapolated networks are more accurate, especially in the early iterations. A straightforward application of our algorithm when training ResNet-152 on ImageNet produces a top-1 test error of 20.88%, improving by 0.8% the reference classification pipeline. Furthermore, the code runs offline in this case, so it never negatively affects performance.
Document type :
Preprints, Working Papers, ...
Liste complète des métadonnées

Cited literature [20 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01799269
Contributor : Damien Scieur <>
Submitted on : Thursday, May 24, 2018 - 3:01:21 PM
Last modification on : Friday, April 19, 2019 - 4:55:14 PM
Document(s) archivé(s) le : Saturday, August 25, 2018 - 2:34:53 PM

File

nonlinear_acceleration_of_deep...
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01799269, version 1

Citation

Damien Scieur, Edouard Oyallon, Alexandre D 'Aspremont, Francis Bach. Nonlinear Acceleration of Deep Neural Networks. 2018. ⟨hal-01799269⟩

Share

Metrics

Record views

149

Files downloads

218