Asynchronous SGD Beats Minibatch SGD Under Arbitrary Delays
Konstantin Mishchenko, Francis Bach, Mathieu Even, Blake Woodworth
Abstract
The existing analysis of asynchronous stochastic gradient descent (SGD) degrades dramatically when any delay is large, giving the impression that performance depends primarily on the delay. On the contrary, we prove much better guarantees for the same asynchronous SGD algorithm regardless of the delays in the gradients, depending instead just on the number of parallel devices used to implement the algorithm. Our guarantees are strictly better than the existing analyses, and we also argue that asynchronous SGD outperforms synchronous minibatch SGD in the settings we consider. For our analysis, we introduce a novel recursion based on ``virtual iterates'' and delay-adaptive stepsizes, which allow us to derive state-of-the-art guarantees for both convex and non-convex objectives.
BibTeX
@inproceedings{
mishchenko2022asynchronous,
title={Asynchronous {SGD} Beats Minibatch {SGD} Under Arbitrary Delays},
author={Konstantin Mishchenko and Francis Bach and Mathieu Even and Blake Woodworth},
booktitle={Advances in Neural Information Processing Systems},
editor={Alice H. Oh and Alekh Agarwal and Danielle Belgrave and Kyunghyun Cho},
year={2022},
url={https://openreview.net/forum?id=4XP0ZuQKXmV}
}