Stochastic Training is Not Necessary for Generalization
Abstract
It is widely believed that the implicit regularization of SGD is fundamental to the impressive generalization behavior we observe in neural networks. In this work, we demonstrate that non-stochastic full-batch training can achieve comparably strong performance to SGD on CIFAR-10 using modern architectures. To this end, we show that the implicit regularization of SGD can be completely replaced with explicit regularization even when comparing against a strong and well-researched baseline. Our observations indicate that the perceived difficulty of full-batch training may be the result of its optimization properties and the disproportionate time and effort spent by the ML community tuning optimizers and hyperparameters for small-batch training.
Cite
@article{arxiv.2109.14119,
title = {Stochastic Training is Not Necessary for Generalization},
author = {Jonas Geiping and Micah Goldblum and Phillip E. Pope and Michael Moeller and Tom Goldstein},
journal= {arXiv preprint arXiv:2109.14119},
year = {2022}
}
Comments
25 pages, 6 figures. Code published at github.com/JonasGeiping/fullbatchtraining. Decompressed version of paper published at ICLR 2022