Lyapunov-Based Optimization of Edge Resources for Energy-Efficient Adaptive Federated Learning

The aim of this paper is to propose a novel dynamic resource allocation strategy for energy-efficient adaptive federated learning at the wireless network edge, with latency and performance guarantees. We consider set devices collecting local data uploading processed information an edge server, which runs stochastic gradient-based algorithms perform continuous adaptation. Hinging on Lyapunov optimization tools, we dynamically optimize radio parameters (e.g., transmitting devices, transmit powers, bits, rates) computation resources CPU cycles server) in order strike best trade-off between power, latency, task. framework admits both model-based implementation, where metrics are available closed-form, data-driven approach, works online estimates interest. method then customized case least mean squares (LMS) estimation, training deep convolutional neural networks. Numerical results illustrate effectiveness our energy-efficient, low-latency, edge.