Parallel Restarted SGD with Faster Convergence and Less Communication: Demystifying Why Model Averaging Works for Deep Learning

TL;DR

This paper provides a rigorous theoretical explanation for why model averaging in distributed deep learning can achieve comparable convergence to parallel mini-batch SGD while significantly reducing communication costs.

Contribution

It offers the first thorough theoretical analysis explaining the effectiveness of model averaging in distributed deep learning training.

Findings

Model averaging achieves similar convergence as mini-batch SGD with less communication.

Theoretical insights clarify why model averaging works well in practice.

Reduces communication overhead without sacrificing convergence speed.

Abstract

In distributed training of deep neural networks, parallel mini-batch SGD is widely used to speed up the training process by using multiple workers. It uses multiple workers to sample local stochastic gradient in parallel, aggregates all gradients in a single server to obtain the average, and update each worker's local model using a SGD update with the averaged gradient. Ideally, parallel mini-batch SGD can achieve a linear speed-up of the training time (with respect to the number of workers) compared with SGD over a single worker. However, such linear scalability in practice is significantly limited by the growing demand for gradient communication as more workers are involved. Model averaging, which periodically averages individual models trained over parallel workers, is another common practice used for distributed training of deep neural networks since (Zinkevich et al. 2010) (McDonald, Hall, and Mann 2010). Compared with parallel mini-batch SGD, the communication overhead of model averaging is significantly reduced. Impressively, tremendous experimental works have verified that model averaging can still achieve a good speed-up of the training time as long as the averaging interval is carefully controlled. However, it remains a mystery in theory why such a simple heuristic works so well. This paper provides a thorough and rigorous theoretical study on why model averaging can work as well as parallel mini-batch SGD with significantly less communication overhead.