Adaptive Quantization of Model Updates for Communication-Efficient Federated Learning

TL;DR

This paper introduces AdaQuantFL, an adaptive quantization method for federated learning that dynamically adjusts quantization levels to reduce communication costs while maintaining model accuracy.

Contribution

It proposes a novel adaptive quantization strategy that changes quantization levels during training to improve communication efficiency in federated learning.

Findings

Reduces communication bits significantly compared to fixed quantization methods.

Maintains comparable training and test accuracy with fewer bits.

Converges faster in terms of communication rounds.

Abstract

Communication of model updates between client nodes and the central aggregating server is a major bottleneck in federated learning, especially in bandwidth-limited settings and high-dimensional models. Gradient quantization is an effective way of reducing the number of bits required to communicate each model update, albeit at the cost of having a higher error floor due to the higher variance of the stochastic gradients. In this work, we propose an adaptive quantization strategy called AdaQuantFL that aims to achieve communication efficiency as well as a low error floor by changing the number of quantization levels during the course of training. Experiments on training deep neural networks show that our method can converge in much fewer communicated bits as compared to fixed quantization level setups, with little or no impact on training and test accuracy.