Non-coherent federated learning improves wireless model updates without channel knowledge
Non-Coherent Over-the-Air Federated Learning: Protocol, Convergence, and Device Scheduling
Information TheoryMachine Learning
Summary
Federated learning lets many devices train a shared model without sending raw data. A big challenge is combining model updates over wireless networks efficiently. The authors propose a method that avoids complex requirements like knowing the exact state of wireless channels, using a simpler detection protocol instead. Their approach converges nearly as fast as the best known methods and smartly chooses which devices participate to speed up learning. Tests on image datasets showed their method performs well in practice.
What this means in practice
- •For wireless network engineers: Implement federated learning systems that aggregate models over wireless channels without needing detailed channel state information, reducing signaling overhead.
- •For edge computing providers: Optimize federated learning across heterogeneous devices by scheduling which ones participate and controlling transmit power to improve training speed under wireless constraints.
- •For smart home device integrators: Enable IoT devices to collaborate on learning local AI models more efficiently over noisy wireless links without complex synchronization.$Commercial implications: Supports development of connected device AI products that lower communication costs and complexity, enhancing user experience in smart homes.
Authors
Haifeng Wen, Nicolò Michelusi, Osvaldo Simeone, Yang Yang, Hong Xing
Abstract
To mitigate the scalability bottleneck in the radio access network (RAN) in federated edge learning (FEEL), over-the-air federated learning (AirFL) exploits waveform superposition over multiple-access channels (MACs) for analog model aggregation. However, coherent AirFL typically relies on stringent PHY-layer conditions such as accurate channel state information (CSI), tight time/frequency synchronization, and frequent transceiver calibration for signal alignment. However, these requirements, if not impossible to be met, incur substantial communication and computation overhead. In this paper, we propose a non-coherent AirFL (NCAirFL) protocol over a broadband single-antenna MAC, leveraging binary dithering, unbiased non-coherent detection, and long-term error feedback to waive the need for instantaneous CSI. For NCAirFL with general smooth non-convex objectives and a constant learning rate, we establish a convergence bound achieving the convergence rate in the same order of $\mathcal{O}(1/\sqrt{T})$ as communication-ideal FedAvg, where $T$ is the total number of communication rounds. To further improve communication efficiency under data and wireless resource heterogeneity, we also derive a lower bound on the expected single-round objective decrease in the global loss conditioned on device scheduling, building upon which a surrogate objective function is obtained for jointly optimal device selection and power control. Experimental results on MNIST and CIFAR-10 corroborate that NCAirFL achieves learning performance close to FedAvg in practical settings, with the proposed device scheduling policy substantially accelerating convergence.