Quantization for Decentralized Learning Under Subspace Constraints - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue IEEE Transactions on Signal Processing Année : 2023

Quantization for Decentralized Learning Under Subspace Constraints

Résumé

In this paper, we consider decentralized optimization problems where agents have individual cost functions to minimize subject to subspace constraints that require the minimizers across the network to lie in low-dimensional subspaces. This constrained formulation includes consensus or single-task optimization as special cases, and allows for more general task relatedness models such as multitask smoothness and coupled optimization. In order to cope with communication constraints, we propose and study an adaptive decentralized strategy where the agents employ differential randomized quantizers to compress their estimates before communicating with their neighbors. The analysis shows that, under some general conditions on the quantization noise, and for sufficiently small step-sizes µ, the strategy is stable both in terms of mean-square error and average bit rate: by reducing µ, it is possible to keep the estimation errors small (on the order of µ) without increasing indefinitely the bit rate as µ → 0 when variablerate quantizers are used. Simulations illustrate the theoretical findings and the effectiveness of the proposed approach, revealing that decentralized learning is achievable at the expense of only a few bits.
Fichier principal
Vignette du fichier
report.pdf (1.87 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04180842 , version 1 (14-08-2023)

Identifiants

Citer

Roula Nassif, Stefan Vlaski, Marco Carpentiero, Vincenzo Matta, Marc Antonini, et al.. Quantization for Decentralized Learning Under Subspace Constraints. IEEE Transactions on Signal Processing, 2023, 71, pp.2320-2335. ⟨10.1109/TSP.2023.3287333⟩. ⟨hal-04180842⟩
20 Consultations
6 Téléchargements

Altmetric

Partager

Gmail Mastodon Facebook X LinkedIn More