Inter-Cell Interference Coordination for AirComp Federated Learning in Cellular Networks
Rongchen Zhang, Na Deng, Haichao Wei · 2023
In large-scale cellular networks, inter-cell interference is a critical factor on the average learning performance in each cell for the over-the-air computation (AirComp) based federated learning (FL). In this paper, we propose a new device selection scheme in which we first establish an interference-reduction region in each cell, then select devices from this region for FL, and the unselected devices in each cell are muted on the resource blocks used for the AirComp to reduce the inter-cell interference. To investigate the performance of the proposed scheme in depth, we consider four types of network topologies with different spatial distribution properties for base stations (BSs) and devices, and build FL experiments based on a neural network and a real dataset. The results show that the proposed scheme efficiently reduces the inter-cell interference while improving the average FL prediction accuracy. And the properties of the spatial distribution of both BSs and devices play a key role on the average FL prediction accuracy.