Pré-Publication, Document De Travail Année : 2025

TOFL: Time Optimized Federated Learning

Résumé

Vehicular networks face threat problems that could harm drivers, passengers, and pedestrians. One possible solution to address these issues is to consider a federated learning-based approach to train models capable of automatically detecting these threats. This approach enhances user security while safeguarding their privacy. Nevertheless, federated learning is particularly sensitive to client delays, especially in vehicular networks where clients constantly move and may experience varying delays or periods of disconnection. In this context, we introduce a client selection strategy designed to minimize the overall training time of a machine-learning model for vehicular threat detection. Our experimental results demonstrate that TOFL, utilizing only 20% of the total available clients, can reduce the time required to reach the target accuracy by up to 50% compared to state-of-the-art approaches while simultaneously decreasing the resource consumption on client devices.

Fichier principal
Vignette du fichier
TOFL__Time_Optimization_Federated_Learning.pdf (1.63 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-05002807 , version 1 (26-03-2025)
hal-05002807 , version 2 (22-01-2026)

Licence

Identifiants

  • HAL Id : hal-05002807 , version 1

Citer

Lucas Airam Castro de Souza, Matteo Sammarco, Nadjib Achir, Miguel Elias Mitre Campista, Luís Henrique Maciel Kosmalski Costa. TOFL: Time Optimized Federated Learning. 2025. ⟨hal-05002807v1⟩
379 Consultations
356 Téléchargements

Partager

  • More