1: \begin{abstract}

2: 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. 

3: We augment the network with Reconfigurable Intelligent Surfaces (RIS) in order to enhance the learning system.

4: We propose a cross-layer algorithm that jointly assigns communication, computation and learning resources. 

5: In particular, we adaptively adjust the number of local steps in conjunction with RIS configuration to boost the learning performance.

6: 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.

7: We provide the convergence analysis for the proposed algorithms and extend the frameworks to personalized learning.

8: Our experimental results demonstrate that the proposed algorithms outperform the state-of-the-art joint communication and learning baselines.

9: \end{abstract}

10: