Communication-Efficient Distributed Learning via Lazily Aggregated Quantized Gradients
This work addresses communication bottlenecks in distributed learning systems, offering a practical solution for large-scale deployments, though it is incremental as it builds on existing gradient-based methods.
The paper tackles the problem of high communication overhead in distributed machine learning by introducing LAQ, a method that combines gradient quantization with adaptive communication skipping, achieving the same linear convergence rate as gradient descent while significantly reducing transmitted bits and communication rounds.
The present paper develops a novel aggregated gradient approach for distributed machine learning that adaptively compresses the gradient communication. The key idea is to first quantize the computed gradients, and then skip less informative quantized gradient communications by reusing outdated gradients. Quantizing and skipping result in `lazy' worker-server communications, which justifies the term Lazily Aggregated Quantized gradient that is henceforth abbreviated as LAQ. Our LAQ can provably attain the same linear convergence rate as the gradient descent in the strongly convex case, while effecting major savings in the communication overhead both in transmitted bits as well as in communication rounds. Empirically, experiments with real data corroborate a significant communication reduction compared to existing gradient- and stochastic gradient-based algorithms.