Scholarlink

<p>Federated Learning enables distributed model training without data centralization, but traditional cloud-based FL suffers from high communication latency, while edge-based FL reduces latency at the cost of model accuracy. To address these challenges, we propose MMVO-SHFL, integrating LSTM-based bandwidth prediction, MAB-driven dynamic client selection, and MVO-guided model parameter optimization. Experiments demonstrate that MMVO-SHFL significantly enhances model convergence speed and accuracy, reduces training time, lowers energy consumption, and improves client participation fairness.</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

MMVO-SHFL: a fair and efficient hierarchical federated learning

  • 期刊论文

  • 发表日期：

    2026年7月24日

查看全文
  • Xia Liu,
  • Jianping Wang,
  • Danyang Chen

摘要

Federated Learning enables distributed model training without data centralization, but traditional cloud-based FL suffers from high communication latency, while edge-based FL reduces latency at the cost of model accuracy. To address these challenges, we propose MMVO-SHFL, integrating LSTM-based bandwidth prediction, MAB-driven dynamic client selection, and MVO-guided model parameter optimization. Experiments demonstrate that MMVO-SHFL significantly enhances model convergence speed and accuracy, reduces training time, lowers energy consumption, and improves client participation fairness.