LGCVMLApr 11, 2020

Detached Error Feedback for Distributed SGD with Random Sparsification

arXiv:2004.05298v313 citations
AI Analysis

This work addresses communication efficiency in large-scale distributed deep learning, offering incremental improvements over existing methods.

The paper tackles the communication bottleneck in distributed SGD by proposing a detached error feedback algorithm that improves convergence and generalization bounds for non-convex problems, showing significant empirical improvements in deep learning experiments.

The communication bottleneck has been a critical problem in large-scale distributed deep learning. In this work, we study distributed SGD with random block-wise sparsification as the gradient compressor, which is ring-allreduce compatible and highly computation-efficient but leads to inferior performance. To tackle this important issue, we improve the communication-efficient distributed SGD from a novel aspect, that is, the trade-off between the variance and second moment of the gradient. With this motivation, we propose a new detached error feedback (DEF) algorithm, which shows better convergence bound than error feedback for non-convex problems. We also propose DEF-A to accelerate the generalization of DEF at the early stages of the training, which shows better generalization bounds than DEF. Furthermore, we establish the connection between communication-efficient distributed SGD and SGD with iterate averaging (SGD-IA) for the first time. Extensive deep learning experiments show significant empirical improvement of the proposed methods under various settings.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes