Quantized Decentralized Stochastic Learning over Directed Graphs

Abstract

We consider a decentralized stochastic learning problem where data points are distributed among computing nodes communicating over a directed graph. As the model size gets large, decentralized learning faces a major bottleneck that is the heavy communication load due to each node transmitting large messages (model updates) to its neighbors. To tackle this bottleneck, we propose the quantized decentralized stochastic learning algorithm over directed graphs that is based on the push-sum algorithm in decentralized consensus optimization. More importantly, we prove that our algorithm achieves the same convergence rates of the decentralized stochastic learning algorithm with exact-communication for both convex and non-convex losses. Numerical evaluations corroborate our main theoretical results and illustrate significant speed-up compared to the exact-communication methods.

fixing typos, minor edits

Cited by in corpus (8)

• Quantization for decentralized learning under subspace constraints
• Quasi-Global Momentum: Accelerating Decentralized Deep Learning on Heterogeneous Data
• Low Precision Decentralized Distributed Training over IID and non-IID Data
• A Hybrid Variance-Reduced Method for Decentralized Stochastic Non-Convex Optimization
• Innovation Compression for Communication-efficient Distributed Optimization with Linear Convergence
• Decentralized Optimization On Time-Varying Directed Graphs Under Communication Constraints
• Gradient-push algorithm for distributed optimization with event-triggered communications
• A unified framework for distributed optimization algorithms over time-varying directed graphs