Scholarlink

<p>As an promising distributed edge learning architecture, federated edge learning (FEEL) facilitates multi-party cooperative model training while preserving data locality. However, the practical deployment of FEEL encounters significant challenges, primarily due to heterogeneity in data distribution and the presence of noisy wireless channels. To tackle these challenges, this paper presents a robust solution for the FEEL system, which integrates a weighted aggregation algorithm based on the index of data completeness to mitigate the adverse effects of heterogeneous data distribution. Concurrently, this solution employs reconfigurable intelligent surface (RIS)-enabled over-the-air computation to optimize communication efficiency and enhance model convergence in heterogeneous wireless environments. Within this framework, we rigorously derive the convergence behavior of the presented FEEL system by taking into account the effects of noisy wireless aggregation and data heterogeneity. Furthermore, we formulate a unified wireless resource optimization problem aimed at minimizing the training loss by jointly optimizing transceiver beamforming and RIS phase shifts. Simulation results demonstrate that the proposed design achieves significant performance improvements compared to several baseline methods.</p>

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

A heterogeneity-aware federated edge learning framework with RIS-empowered wireless aggregation

  • 期刊论文

  • 发表日期：

    2025年12月3日

查看全文
  • Tianli Song,
  • Shaoyi Han,
  • Shengliang Wu,
  • Heju Li,
  • Xin He

摘要

As an promising distributed edge learning architecture, federated edge learning (FEEL) facilitates multi-party cooperative model training while preserving data locality. However, the practical deployment of FEEL encounters significant challenges, primarily due to heterogeneity in data distribution and the presence of noisy wireless channels. To tackle these challenges, this paper presents a robust solution for the FEEL system, which integrates a weighted aggregation algorithm based on the index of data completeness to mitigate the adverse effects of heterogeneous data distribution. Concurrently, this solution employs reconfigurable intelligent surface (RIS)-enabled over-the-air computation to optimize communication efficiency and enhance model convergence in heterogeneous wireless environments. Within this framework, we rigorously derive the convergence behavior of the presented FEEL system by taking into account the effects of noisy wireless aggregation and data heterogeneity. Furthermore, we formulate a unified wireless resource optimization problem aimed at minimizing the training loss by jointly optimizing transceiver beamforming and RIS phase shifts. Simulation results demonstrate that the proposed design achieves significant performance improvements compared to several baseline methods.