signSGD: Compressed Optimisation for Non-Convex Problems

Abstract

Training large neural networks requires distributing learning across multiple workers, where the cost of communicating gradients can be a significant bottleneck. signSGD alleviates this problem by transmitting just the sign of each minibatch stochastic gradient. We prove that it can get the best of both worlds: compressed gradients and SGD-level convergence rate. The relative ${\ell }_1/{\ell }_2$ geometry of gradients, noise and curvature informs whether signSGD or SGD is theoretically better suited to a particular problem. On the practical side we find that the momentum counterpart of signSGD is able to match the accuracy and convergence speed of Adam on deep Imagenet models. We extend our theory to the distributed setting, where the parameter server uses majority vote to aggregate gradient signs from each worker enabling 1-bit compression of worker-server communication in both directions. Using a theorem by Gauss we prove that majority vote can achieve the same reduction in variance as full precision distributed SGD. Thus, there is great promise for sign-based optimisation schemes to achieve fast communication and fast convergence. Code to reproduce experiments is to be found at https://github.com/jxbz/signSGD .

Cited by in corpus (13)

• Large Batch Optimization for Deep Learning: Training BERT in 76 minutes
• Near-Optimal Sparse Allreduce for Distributed Deep Learning
• Adam revisited: a weighted past gradients perspective
• DataLens: Scalable Privacy Preserving Training via Gradient Compression and Aggregation
• Communication-Efficient Edge AI: Algorithms and Systems
• Taming Momentum in a Distributed Asynchronous Environment
• Gradient Descent with Compressed Iterates
• BlockFLA: Accountable Federated Learning via Hybrid Blockchain Architecture
• Beyond Throughput and Compression Ratios: Towards High End-to-end Utility of Gradient Compression
• Federated Learning in Adversarial Settings
• Sign Bit is Enough: A Learning Synchronization Framework for Multi-hop All-reduce with Ultimate Compression
• Constrained Differentially Private Federated Learning for Low-bandwidth Devices
• Doing More by Doing Less: How Structured Partial Backpropagation Improves Deep Learning Clusters