Biased Over-the-Air Federated Learning under Wireless Heterogeneity

TL;DR

This paper explores biased over-the-air federated learning in wireless heterogeneous environments, proposing pre-scaler designs that balance bias and variance to improve convergence and performance over traditional zero-bias methods.

Contribution

It introduces a new bias-variance trade-off framework for OTA-FL pre-scaler design under wireless heterogeneity, with derived bounds and practical solutions.

Findings

Biased pre-scalers can outperform zero-bias schemes in heterogeneous settings.

Minimizing variance with a small bias improves FL convergence.

Numerical results confirm the effectiveness of the proposed approach.

Abstract

Recently, Over-the-Air (OTA) computation has emerged as a promising federated learning (FL) paradigm that leverages the waveform superposition properties of the wireless channel to realize fast model updates. Prior work focused on the OTA device ``pre-scaler" design under \emph{homogeneous} wireless conditions, in which devices experience the same average path loss, resulting in zero-bias solutions. Yet, zero-bias designs are limited by the device with the worst average path loss and hence may perform poorly in \emph{heterogeneous} wireless settings. In this scenario, there may be a benefit in designing \emph{biased} solutions, in exchange for a lower variance in the model updates. To optimize this trade-off, we study the design of OTA device pre-scalers by focusing on the OTA-FL convergence. We derive an upper bound on the model ``optimality error", which explicitly captures the effect of bias and variance in terms of the choice of the pre-scalers. Based on this bound, we identify two solutions of interest: minimum noise variance, and minimum noise variance zero-bias solutions. Numerical evaluations show that using OTA device pre-scalers that minimize the variance of FL updates, while allowing a small bias, can provide high gains over existing schemes.