On the Linear Speedup Analysis of Communication Efficient Momentum SGD for Distributed Non-Convex Optimization

Abstract

Recent developments on large-scale distributed machine learning applications, e.g., deep neural networks, benefit enormously from the advances in distributed non-convex optimization techniques, e.g., distributed Stochastic Gradient Descent (SGD). A series of recent works study the linear speedup property of distributed SGD variants with reduced communication. The linear speedup property enable us to scale out the computing capability by adding more computing nodes into our system. The reduced communication complexity is desirable since communication overhead is often the performance bottleneck in distributed systems. Recently, momentum methods are more and more widely adopted in training machine learning models and can often converge faster and generalize better. For example, many practitioners use distributed SGD with momentum to train deep neural networks with big data. However, it remains unclear whether any distributed momentum SGD possesses the same linear speedup property as distributed SGD and has reduced communication complexity. This paper fills the gap by considering a distributed communication efficient momentum SGD method and proving its linear speedup property.

A short version of this paper is accepted to ICML 2019

Cited by in corpus (9)

• Variance Reduced Local SGD with Lower Communication Complexity
• SlowMo: Improving Communication-Efficient Distributed SGD with Slow Momentum
• Privacy-preserving Federated Brain Tumour Segmentation
• Secure Federated Submodel Learning
• Learn Electronic Health Records by Fully Decentralized Federated Learning
• SPARQ-SGD: Event-Triggered and Compressed Communication in Decentralized Stochastic Optimization
• Byzantine Resilient Non-Convex SVRG with Distributed Batch Gradient Computations
• Distributed Optimization over Block-Cyclic Data
• Adaptive Serverless Learning