Scholarlink

Federated Learning (FL) is a pivotal technology in the era of big data and Artificial Intelligence, addressing crucial aspects of data privacy preservation and alleviating the necessity to transfer massive datasets. Unlike conventional central training, FL involves collaborative model training through exchanged parameter updates, but challenges arise due to communication and computation overhead, particularly in applications like the Internet of Things (IoT). This paper presents an analytical study of FedProx, a framework specifically designed to address heterogeneity in federated networks. FedProx extends the state-of-the-art method FedAvg by providing convergence guarantees for non-identically distributed data and accommodating variable workloads on each device. This research endeavors to provide a comprehensive analysis of FedProx, exploring its effectiveness and robustness in heterogeneous federated environments, and a deeper insights into strengths and limitations.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Advancements in Federated Learning: A Analysis Study of Federated Optimization in Heterogeneous Networks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Alami Aroussi Said,
  • Silkan Hassan,
  • Elouadrhiri Ahmed,
  • Hadni Meryeme

摘要

Federated Learning (FL) is a pivotal technology in the era of big data and Artificial Intelligence, addressing crucial aspects of data privacy preservation and alleviating the necessity to transfer massive datasets. Unlike conventional central training, FL involves collaborative model training through exchanged parameter updates, but challenges arise due to communication and computation overhead, particularly in applications like the Internet of Things (IoT). This paper presents an analytical study of FedProx, a framework specifically designed to address heterogeneity in federated networks. FedProx extends the state-of-the-art method FedAvg by providing convergence guarantees for non-identically distributed data and accommodating variable workloads on each device. This research endeavors to provide a comprehensive analysis of FedProx, exploring its effectiveness and robustness in heterogeneous federated environments, and a deeper insights into strengths and limitations.