English

L-FGADMM: Layer-Wise Federated Group ADMM for Communication Efficient Decentralized Deep Learning

Machine Learning 2020-07-07 v2 Distributed, Parallel, and Cluster Computing Networking and Internet Architecture Machine Learning

Abstract

This article proposes a communication-efficient decentralized deep learning algorithm, coined layer-wise federated group ADMM (L-FGADMM). To minimize an empirical risk, every worker in L-FGADMM periodically communicates with two neighbors, in which the periods are separately adjusted for different layers of its deep neural network. A constrained optimization problem for this setting is formulated and solved using the stochastic version of GADMM proposed in our prior work. Numerical evaluations show that by less frequently exchanging the largest layer, L-FGADMM can significantly reduce the communication cost, without compromising the convergence speed. Surprisingly, despite less exchanged information and decentralized operations, intermittently skipping the largest layer consensus in L-FGADMM creates a regularizing effect, thereby achieving the test accuracy as high as federated learning (FL), a baseline method with the entire layer consensus by the aid of a central entity.

Keywords

Cite

@article{arxiv.1911.03654,
  title  = {L-FGADMM: Layer-Wise Federated Group ADMM for Communication Efficient Decentralized Deep Learning},
  author = {Anis Elgabli and Jihong Park and Sabbir Ahmed and Mehdi Bennis},
  journal= {arXiv preprint arXiv:1911.03654},
  year   = {2020}
}

Comments

6 pages; 4 figures; presented at IEEE WCNC'2020

R2 v1 2026-06-23T12:10:09.380Z