Saved in:
Bibliographic Details
Main Authors: Campbell, Andrew, Liu, Hang, Woldemariam, Leah, Scaglione, Anna
Format: Preprint
Published: 2023
Subjects:
Online Access:https://arxiv.org/abs/2311.05760
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929375672270848
author Campbell, Andrew
Liu, Hang
Woldemariam, Leah
Scaglione, Anna
author_facet Campbell, Andrew
Liu, Hang
Woldemariam, Leah
Scaglione, Anna
contents Recent research highlights frequent model communication as a significant bottleneck to the efficiency of decentralized machine learning (ML), especially for large-scale and over-parameterized neural networks (NNs). To address this, we present Malcom-PSGD, a novel decentralized ML algorithm that combines gradient compression techniques with model sparsification. We promote model sparsity by adding $\ell_1$ regularization to the objective and present a decentralized proximal SGD method for training. Our approach employs vector source coding and dithering-based quantization for the compressed gradient communication of sparsified models. Our analysis demonstrates that Malcom-PSGD achieves a convergence rate of $\mathcal{O}(1/\sqrt{t})$ with respect to the iterations $t$, assuming a constant consensus and learning rate. This result is supported by our proof for the convergence of non-convex compressed Proximal SGD methods. Additionally, we conduct a bit analysis, providing a closed-form expression for the communication costs associated with Malcom-PSGD. Numerical results verify our theoretical findings and demonstrate that our method reduces communication costs by approximately $75\%$ when compared to the state-of-the-art.
format Preprint
id arxiv_https___arxiv_org_abs_2311_05760
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Compressed and Sparse Models for Non-Convex Decentralized Learning
Campbell, Andrew
Liu, Hang
Woldemariam, Leah
Scaglione, Anna
Machine Learning
Distributed, Parallel, and Cluster Computing
Data Structures and Algorithms
Multiagent Systems
Optimization and Control
68W15, 68W10, 68W40, 90C06, 90C35, 90C25
G.1.6; F.2.1; E.4
Recent research highlights frequent model communication as a significant bottleneck to the efficiency of decentralized machine learning (ML), especially for large-scale and over-parameterized neural networks (NNs). To address this, we present Malcom-PSGD, a novel decentralized ML algorithm that combines gradient compression techniques with model sparsification. We promote model sparsity by adding $\ell_1$ regularization to the objective and present a decentralized proximal SGD method for training. Our approach employs vector source coding and dithering-based quantization for the compressed gradient communication of sparsified models. Our analysis demonstrates that Malcom-PSGD achieves a convergence rate of $\mathcal{O}(1/\sqrt{t})$ with respect to the iterations $t$, assuming a constant consensus and learning rate. This result is supported by our proof for the convergence of non-convex compressed Proximal SGD methods. Additionally, we conduct a bit analysis, providing a closed-form expression for the communication costs associated with Malcom-PSGD. Numerical results verify our theoretical findings and demonstrate that our method reduces communication costs by approximately $75\%$ when compared to the state-of-the-art.
title Compressed and Sparse Models for Non-Convex Decentralized Learning
topic Machine Learning
Distributed, Parallel, and Cluster Computing
Data Structures and Algorithms
Multiagent Systems
Optimization and Control
68W15, 68W10, 68W40, 90C06, 90C35, 90C25
G.1.6; F.2.1; E.4
url https://arxiv.org/abs/2311.05760