RIS-Assisted Wireless FL Optimizes Convergence and Latency

Liwei Wang, Wen Chen, Jun Li, Qingqing Wu, Ming Ding, Xusheng Zhu, Qiong Wu· July 23, 2026 View original

Summary

This paper proposes an adaptive modulation and resource allocation scheme for RIS-assisted wireless Federated Learning (FL) to address training latency and convergence issues under unreliable transmission. It derives a convergence-related upper bound considering symbol errors and formulates a joint optimization problem, showing improved convergence and accuracy.

This research addresses the significant challenges of training latency and degraded convergence in Federated Learning (FL) when deployed over wireless networks, especially in environments with blocked propagation. While Reconfigurable Intelligent Surfaces (RISs) can enhance communication reliability, existing studies often overlook the trade-off between learning convergence and communication delay, particularly concerning modulation-dependent transmission errors. The paper introduces a wireless FL system that operates with RIS assistance in blocked-link scenarios. It focuses on optimizing adaptive modulation and sub-channel allocation to achieve convergence-latency aware communication. A key contribution is the derivation of a convergence-related upper bound that quantifies the impact of symbol error rate (SER) on FL loss decay. Based on this, a complex mixed-integer nonlinear programming (MINLP) problem is formulated and solved using a low-complexity hybrid alternating optimization framework. Extensive experiments across various datasets demonstrate that the proposed scheme consistently delivers faster convergence and higher test accuracy compared to other adaptive communication methods, particularly in challenging wireless conditions and for complex tasks.

Why it matters

Professionals deploying Federated Learning in real-world wireless environments can use this research to design more efficient and reliable systems that balance communication latency with model convergence and accuracy.

How to implement this in your domain

  1. 1Evaluate the feasibility of integrating Reconfigurable Intelligent Surfaces (RISs) into your wireless FL infrastructure.
  2. 2Implement adaptive modulation and sub-channel allocation strategies based on the derived convergence-latency trade-offs.
  3. 3Develop mechanisms to characterize and account for symbol error rates (SER) in your FL loss decay models.
  4. 4Explore the proposed hybrid alternating optimization framework for joint convergence-latency optimization in your FL deployments.

Who benefits

TelecommunicationsIoTSmart CitiesHealthcareAutomotive

Key takeaways

  • RIS-assisted wireless FL can significantly improve training latency and convergence.
  • The research characterizes the impact of symbol errors on FL loss decay.
  • A new optimization scheme balances communication delay and learning convergence.
  • The proposed method achieves faster convergence and higher accuracy in challenging wireless scenarios.

Original post by Liwei Wang, Wen Chen, Jun Li, Qingqing Wu, Ming Ding, Xusheng Zhu, Qiong Wu

"arXiv:2607.19759v1 Announce Type: new Abstract: Federated learning (FL) over wireless networks suffers from significant training latency and degraded convergence due to unreliable wireless transmission, especially under blocked propagation environments. Although reconfigurable in…"

View on X

Originally posted by Liwei Wang, Wen Chen, Jun Li, Qingqing Wu, Ming Ding, Xusheng Zhu, Qiong Wu on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses