Work

Decentralised convex optimisation with probability-proportional-to-size quantization

Year: 2025

Type: article

Abstract: Communication is one of the bottlenecks of distributed optimisation and learning. To overcome this bottleneck, we propose a novel quantization method that transforms a vector into a sample of components' indices drawn from a categorical distribution with probabilities proportional to values at those components. Then, we propose a primal and a primal-dual accelerated stochastic gradient methods that use our proposed quantization, and derive their convergence rates in terms of probabilities of large deviations. We focus on affine-constrained convex optimisation and its application to decentralised distributed optimisation problems. To illustrate the work of our algorithm, we apply it to the decentralised computation of semi-discrete entropy regularized Wasserstein barycentre's.

Source: EURO Journal on Computational Optimization

DOI: 10.1016/j.ejco.2025.100113

Authors: D.A. Pasechnyuk, P. Dvurechensky, C.A. Uribe, A.V. Gasnikov

Institutions: Mohamed bin Zayed University of Artificial Intelligence, Weierstrass Institute for Applied Analysis and Stochastics, Rice University, Moscow Institute of Physics and Technology, Innopolis University

Language: English

Topic: Distributed Control Multi-Agent Systems

Subfield: Computer Networks and Communications

Field: Computer Science

Domain: Physical Sciences

Open Access status: gold

Estimated APC paid: $1,400