Abstract:Over-the-air federated learning (OTA-FL) unifies communication and model aggregation by leveraging the inherent superposition property of the wireless medium. This strategy can enable scalable and bandwidth-efficient learning via simultaneous transmission of model updates using the same frequency resources, if care is exercised to design the physical layer jointly with learning. In this paper, a federated learning system facilitated by a heterogeneous edge-intelligent network is considered. The edge users (clients) have differing user resources and non-i.i.d. local dataset distributions. A general non-convex learning objective is considered for the model training task(s) at hand. We augment the network with Reconfigurable Intelligent Surfaces (RIS) in order to enhance the learning system. We propose a cross-layer algorithm that jointly assigns communication, computation and learning resources. In particular, we adaptively adjust the number of local steps in conjunction with RIS configuration to boost the learning performance. Our system model considers channel noise and channel estimation errors in both the uplink (model updates) and downlink (global model broadcast), employing dynamic power control for both. We provide the convergence analysis for the proposed algorithms and extend the frameworks to personalized learning. Our experimental results demonstrate that the proposed algorithms outperform the state-of-the-art joint communication and learning baselines.

What problem does this paper attempt to address?

The main problems that this paper attempts to solve are as follows: In the Federated Learning (FL) framework, how to enhance the learning performance of edge devices by introducing Reconfigurable Intelligent Surfaces (RIS), and at the same time optimize the allocation of communication, computing and learning resources to deal with problems such as wireless channel noise, channel estimation errors and data heterogeneity. Specifically: 1. **Improve learning performance**: By introducing RIS to improve the wireless propagation environment, thereby improving model aggregation and overall learning performance. 2. **Optimize resource allocation**: A cross - layer algorithm is proposed to jointly optimize communication, computing and learning resources, including dynamically adjusting RIS phase shift, local update steps and transmission power, in order to mitigate the impact of time - varying imperfect channel state information (CSI) and data/system heterogeneity. 3. **Deal with channel noise and estimation errors**: Consider the noise and channel estimation errors in the uplink and downlink, and use dynamic power control to deal with these problems. 4. **Personalized federated learning**: Extend the proposed framework to Personalized Federated Learning (PFL) and verify the potential of the individual RIS model in personalized learning. The main contributions of the paper include: - Propose the ROAR - Fed algorithm, which can adaptively adjust RIS phase shift, local update steps and transmission power to alleviate the impact of channel noise and data heterogeneity. - Provide convergence analysis, revealing the impact of imperfect CSI and communication noise on learning and communication errors. - Verify the superior performance of the proposed algorithm in heterogeneous networks through experiments and extend it to the personalized federated learning scenario. In terms of formulas, the formulas involved in the paper are as follows: - Global model update formula: \[ w^{t + 1}=w^t+\frac{1}{\beta_u^t}\sum_{i = 1}^m h_{u,i}^t x_i^t+\tilde{z}_u^t \] - Dynamic power control parameter design: \[ \beta_i^t=\frac{\beta_u^t\alpha_i}{\tau_i^t\hat{h}_{u,i}^t} \] - Statistical error and channel estimation error in convergence analysis: \[ L\eta\sigma^2\frac{1}{T}\sum_{t = 0}^{T - 1}\sum_{i = 1}^m\alpha_i^2E_t\left\|\frac{h_{u,i}^t}{\hat{h}_{u,i}^t}\right\|^2\leq L\eta\sigma^2\sum_{i = 1}^m\alpha_i^2(1 + C) \] These formulas show the key technical details in the paper and ensure the correctness and readability of the formulas.

Adaptive Personalized Over-the-Air Federated Learning with Reflecting Intelligent Surfaces

RIS-Assisted Over-the-Air Adaptive Federated Learning with Noisy Downlink

Personalized Over-the-Air Federated Learning with Personalized Reconfigurable Intelligent Surfaces

ROAR-Fed: RIS-Assisted Over-the-Air Adaptive Resource Allocation for Federated Learning

Reconfigurable Intelligent Surface Enabled Federated Learning: A Unified Communication-Learning Design Approach

Federated Learning in Multi-RIS-Aided Systems

Deep Learning Based Coded Over-the-Air Computation for Personalized Federated Learning

Federated Spectrum Learning for Reconfigurable Intelligent Surfaces-Aided Wireless Edge Networks

Differentially Private Federated Learning via Reconfigurable Intelligent Surface

Asynchronous Federated Learning over Wireless Communication Networks.

Hierarchical Federated Learning with Adaptive Clustering on Non-IID Data

Over-The-Air Federated Learning: Status Quo, Open Challenges, and Future Directions

Over-the-air Learning Rate Optimization for Federated Learning

Over-the-Air Federated Learning with Joint Adaptive Computation and Power Control

Performance-Oriented Design for Intelligent Reflecting Surface Assisted Federated Learning

Reconfigurable Intelligent Surface Empowered Over-the-Air Federated Edge Learning

Deep Reinforcement Learning for Multi-Functional RIS-Aided Over-the-Air Federated Learning in Internet of Robotic Things

Over-the-Air Federated Learning and Non-Orthogonal Multiple Access Unified by Reconfigurable Intelligent Surface.

Empowering Over-the-Air Personalized Federated Learning via RIS

Federated Learning with Unsourced Random Access.

STAR-RIS Integrated Nonorthogonal Multiple Access and Over-the-Air Federated Learning: Framework, Analysis, and Optimization