Adaptive Federated Learning Over the Air

arXiv:2403.06528v1 Announce Type: new
Abstract: We propose a federated version of adaptive gradient methods, particularly AdaGrad and Adam, within the framework of over-the-air model training. This approach capitalizes on the inherent superposition property of wireless channels, facilitating fast and scalable parameter aggregation. Meanwhile, it enhances the robustness of the model training process by dynamically adjusting the stepsize in accordance with the global gradient update. We derive the convergence rate of the training algorithms, encompassing the effects of channel fading and …

abstract adam aggregation arxiv channels cs.it cs.lg cs.ni federated learning framework gradient math.it process property robustness scalable superposition training type wireless