Academic Journal

In-depth Analysis of Low-rank Matrix Factorisation in a Federated Setting

Bibliographic Details
Title: In-depth Analysis of Low-rank Matrix Factorisation in a Federated Setting
Authors: Philippenko, Constantin, Scaman, Kevin, Massoulié, Laurent
Contributors: Philippenko, Constantin
Source: Proceedings of the AAAI Conference on Artificial Intelligence. 39:19904-19912
Publication Status: Preprint
Publisher Information: Association for the Advancement of Artificial Intelligence (AAAI), 2025.
Publication Year: 2025
Subject Terms: Optimization, Matrix factorisation, Machine Learning, FOS: Computer and information sciences, Optimization and Control (math.OC), Optimization and Control, Federated learning, FOS: Mathematics, [INFO.INFO-LG] Computer Science [cs]/Machine Learning [cs.LG], [STAT.ML] Statistics [stat]/Machine Learning [stat.ML], Machine Learning (cs.LG)
Description: This work presents a novel approach to low-rank matrix factorization in a federated learning context, where multiple clients collaboratively solve a matrix decomposition problem without sharing their local data. The algorithm introduces a power initialization technique for the global factorization matrix and combines it with local gradient descent updates to achieve strong theoretical and practical guarantees. Considering this power initialization, we rewrite the previous smooth non-convex problem into a smooth strongly-convex problem that we solve using a parallel Nesterov gradient descent potentially requiring a single step of communication at the initialization step. We provide a linear rate of convergence of the excess loss, our results improve the rates of convergence given in the literature. We provide an upper bound on the Frobenius-norm error of reconstruction under the power initialization strategy. We complete our analysis with experiments on both synthetic and real data.
Document Type: Article
Conference object
File Description: application/pdf
ISSN: 2374-3468
2159-5399
DOI: 10.1609/aaai.v39i19.34192
DOI: 10.48550/arxiv.2409.08771
Access URL: http://arxiv.org/abs/2409.08771
https://hal.science/hal-04893622v2
Rights: CC BY
Accession Number: edsair.doi.dedup.....43f1eae6b3e8dfb7e16e7bd3ff8606d2
Database: OpenAIRE
Description
ISSN:23743468
21595399
DOI:10.1609/aaai.v39i19.34192