M22: A Communication-Efficient Algorithm for Federated Learning Inspired by Rate-Distortion

Yangyi Liu, Stefano Rini, Sadaf Salehkalaibar, Jun Chen

Research output: Contribution to journalArticlepeer-review

Abstract

In federated learning (FL), the communication constraint between the remote clients and the Parameter Server (PS) is a crucial bottleneck. For this reason, model updates must be compressed so as to minimize the loss in accuracy resulting from the communication constraint. This paper proposes &#x201C;M-<italic>magnitude weighted L2 distortion</italic> + 2 <italic>degrees of freedom</italic>&#x201D; (M22) algorithm, a rate-distortion inspired approach to gradient compression for federated training of deep neural networks (DNNs). In particular, we propose a family of distortion measures between the original gradient and the reconstruction we referred to as &#x201C;<italic>M</italic>-magnitude weighted <italic>L</italic>2&#x201D; distortion, and we assume that gradient updates follow an i.i.d. distribution &#x2013; generalized normal or Weibull, which have two degrees of freedom. In both the distortion measure and the gradient distribution, there is one free parameter for each that can be fitted as a function of the iteration number. Given a choice of gradient distribution and distortion measure, we design the quantizer to minimize the expected distortion in gradient reconstruction. To measure the gradient compression performance under a communication constraint, we define the <italic>per-bit accuracy</italic> as the optimal improvement in accuracy that one bit of communication brings to the centralized model over the training period. Using this performance measure, we systematically benchmark the choice of gradient distribution and distortion measure. We provide substantial insights on the role of these choices and argue that significant performance improvements can be attained using such a rate-distortion inspired compressor.

Original languageEnglish
Pages (from-to)1
Number of pages1
JournalIEEE Transactions on Communications
DOIs
StateAccepted/In press - 2023

Keywords

  • Compression algorithms
  • Distortion
  • Distortion measurement
  • DNN gradient modeling
  • Federated learning
  • Federated learning
  • Gradient compression
  • Gradient quantization
  • Gradient sparsification
  • Quantization (signal)
  • Rate-distortion
  • Training

Fingerprint

Dive into the research topics of 'M22: A Communication-Efficient Algorithm for Federated Learning Inspired by Rate-Distortion'. Together they form a unique fingerprint.

Cite this