Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fed-PELAD: Communication-Efficient Federated Learning for Massive MIMO CSI Feedback with Personalized Encoders and a LoRA-Adapted Shared Decoder

Yixiang Zhou, Tong Wu, Meixia Tao, Jianhua Mo

TL;DR

This work tackles the high communication and privacy costs of CSI feedback in FDD massive MIMO by proposing Fed-PELAD, a federated learning framework with device-specific encoders and a LoRA-adapted shared decoder. By keeping encoders local and exchanging compact LoRA adapters, it achieves substantial uplink savings while handling data heterogeneity. An alternating-freeze strategy with learning-rate calibration stabilizes convergence under heterogeneous CSI distributions. Experiments on 3GPP-channel models show Fed-PELAD reduces uplink cost to 42.97% of full-model FedAvg and delivers up to 1.28 dB NMSE improvement, with AF further enhancing robustness and a half-budget variant offering strong accuracy with minimal bandwidth.

Abstract

This paper addresses the critical challenges of communication overhead, data heterogeneity, and privacy in deep learning for channel state information (CSI) feedback in massive MIMO systems. To this end, we propose Fed-PELAD, a novel federated learning framework that incorporates personalized encoders and a LoRA-adapted shared decoder. Specifically, personalized encoders are trained locally on each user equipment (UE) to capture device-specific channel characteristics, while a shared decoder is updated globally via the coordination of the base station (BS) by using Low-Rank Adaptation (LoRA). This design ensures that only compact LoRA adapter parameters instead of full model updates are transmitted for aggregation. To further enhance convergence stability, we introduce an alternating freezing strategy with calibrated learning-rate ratio during LoRA aggregation. Extensive simulations on 3GPP-standard channel models demonstrate that Fed-PELAD requires only 42.97\% of the uplink communication cost compared to conventional methods while achieving a performance gain of 1.2 dB in CSI feedback accuracy under heterogeneous conditions.

Fed-PELAD: Communication-Efficient Federated Learning for Massive MIMO CSI Feedback with Personalized Encoders and a LoRA-Adapted Shared Decoder

TL;DR

This work tackles the high communication and privacy costs of CSI feedback in FDD massive MIMO by proposing Fed-PELAD, a federated learning framework with device-specific encoders and a LoRA-adapted shared decoder. By keeping encoders local and exchanging compact LoRA adapters, it achieves substantial uplink savings while handling data heterogeneity. An alternating-freeze strategy with learning-rate calibration stabilizes convergence under heterogeneous CSI distributions. Experiments on 3GPP-channel models show Fed-PELAD reduces uplink cost to 42.97% of full-model FedAvg and delivers up to 1.28 dB NMSE improvement, with AF further enhancing robustness and a half-budget variant offering strong accuracy with minimal bandwidth.

Abstract

This paper addresses the critical challenges of communication overhead, data heterogeneity, and privacy in deep learning for channel state information (CSI) feedback in massive MIMO systems. To this end, we propose Fed-PELAD, a novel federated learning framework that incorporates personalized encoders and a LoRA-adapted shared decoder. Specifically, personalized encoders are trained locally on each user equipment (UE) to capture device-specific channel characteristics, while a shared decoder is updated globally via the coordination of the base station (BS) by using Low-Rank Adaptation (LoRA). This design ensures that only compact LoRA adapter parameters instead of full model updates are transmitted for aggregation. To further enhance convergence stability, we introduce an alternating freezing strategy with calibrated learning-rate ratio during LoRA aggregation. Extensive simulations on 3GPP-standard channel models demonstrate that Fed-PELAD requires only 42.97\% of the uplink communication cost compared to conventional methods while achieving a performance gain of 1.2 dB in CSI feedback accuracy under heterogeneous conditions.

Paper Structure

This paper contains 13 sections, 11 equations, 2 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model
  3. The proposed Fed-PELAD Algorithm
  4. FEEL-based Personalized Encoders and a Shared Decoder
  5. Alternating-Freeze Aggregation for the Decoder
  6. Simulation Results
  7. Simulation Settings
  8. Datasets and Model
  9. Baselines and LoRA-adapted Layers
  10. Federated Training Protocol, Hyperparameters and Metrics
  11. Performance Evaluation
  12. Impact of Hyperparameters on Fed-PELAD
  13. Conclusion

Figures (2)

  • Figure 1: Fed-PELAD Framework: Personalized Encoders and a LoRA-Adapted Shared Decoder with Federated Adapter Aggregation.
  • Figure 2: NMSE (dB) Performance of Fed-PELAD and Baselines vs. Cumulative Uplink Cost (Million parameters) under UMi scenario with four UEs