Over-the-Air Federated Learning: Rethinking Edge AI Through Signal Processing

The article presents Over-the-Air Federated Learning (AirFL), a novel approach that integrates wireless signal processing with distributed machine learning to enhance AI scalability at the edge. AirFL utilizes wireless superposition to aggregate local model updates into an analog signal, significantly reducing communication latency, bandwidth, and energy consumption. The paper categorizes existing AirFL schemes into three classes—CSIT-aware, blind, and weighted—while discussing their performance trade-offs, complexities, and potential applications in practical wireless edge-AI systems, providing insights for practitioners in optimizing federated learning in constrained environments.