A Superquantile Approach to Federated Learning with Heterogeneous Devices - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2021

A Superquantile Approach to Federated Learning with Heterogeneous Devices

Résumé

We present a federated learning framework that allows one to handle heterogeneous client devices that may not conform to the population data distribution. The proposed approach hinges upon a parameterized superquantile-based objective, where the parameter ranges over levels of conformity. We introduce a stochastic optimization algorithm compatible with secure aggregation, which interleaves device filtering steps with federated averaging steps. We conclude with numerical experiments with neural networks on computer vision and natural language processing data.
Fichier principal
Vignette du fichier
ciss_paper.pdf (395.12 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03750727 , version 1 (12-08-2022)

Identifiants

Citer

Yassine Laguel, Krishna Pillutla, Jerôme Malick, Zaid Harchaoui. A Superquantile Approach to Federated Learning with Heterogeneous Devices. 55th Annual Conference on Information Sciences and Systems (CISS), Mar 2021, Baltimore, United States. ⟨10.1109/CISS50987.2021.9400318⟩. ⟨hal-03750727⟩

Collections

UGA CNRS MIAI ANR
10 Consultations
49 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More