Recently, a new machine learning paradigm, namely Federated Learning (FL), has received considerable attentions in vehicular networks (VANETs). Different from the centralized deployment, the decentralized FL enables vehicle nodes to use their own data to learn a global model collaboratively, thus can preserve the privacy of the raw data while eliminating data traffic jam encountered by the central server. However, the environmental dynamics as well as the instability of communication links in VANETs bring in new challenges in designing consensus strategies in decentralized FL. In this work, we investigate an online link reliability sampling strategy, based on which the weight matrix employed in the consensus process can be determined on-the-fly. This method does NOT require any knowledge of the network topology and is totally parameter free. We further provide a distributed stochastic gradient descent (DSGD) algorithm, termed as Samp-DSGD, for deploying decentralized FL over unreliable communication links. By developing mathematical models, we analyze the convergence rate of the Samp-DSGD algorithm. Via conducting numerical experiments, we showcase the performance efficiency of Samp-DSGD in comparison with state-of-the-art algorithms under various setups.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Samp-DSGD: Link Reliability Sampling Based Federated Learning for Vehicular Networks


    Contributors:


    Publication date :

    2024-09-01


    Size :

    5339488 byte




    Type of media :

    Article (Journal)


    Type of material :

    Electronic Resource


    Language :

    English



    Federated Learning in Vehicular Networks

    Elbir, Ahmet M. / Soner, Burak / Coleri, Sinem et al. | IEEE | 2022


    Blockchain-Enabled Federated Learning Approach for Vehicular Networks

    Sultana, Shirin / Hossain, Jahin / Billah, Maruf et al. | ArXiv | 2023

    Free access

    Federated Learning for Anomaly Detection in Vehicular Networks

    Tham, Chen-Khong / Yang, Lu / Khanna, Akshit et al. | IEEE | 2023


    Asynchronous Federated Learning for Edge-assisted Vehicular Networks

    Wang, Siyuan / Wu, Qiong / Fan, Qiang et al. | ArXiv | 2022

    Free access

    RCFL: Redundancy-Aware Collaborative Federated Learning in Vehicular Networks

    Hui, Yilong / Hu, Jie / Cheng, Nan et al. | IEEE | 2024