Engineering and Technology

Balancing Energy Efficiency and Distributional Robustness in Over-the-Air Federated Learning

M. Badi, C. B. Issaid, et al.

Discover a groundbreaking method for enhancing energy efficiency in federated learning through innovative over-the-air computation, presented by Mohamed Badi, Chaouki Ben Issaid, Anis Elgabli, and Mehdi Bennis. This research tackles critical challenges in distributed learning, achieving remarkable energy savings while maintaining robustness against data variability.

00:00

Playback language: English

Index

Abstract

This paper introduces a novel approach for energy-efficient and distributionally robust federated learning (FL) using over-the-air computation (AirComp). It addresses challenges posed by energy consumption, bandwidth limitations, latency, and data heterogeneity in distributed learning. A novel client selection method is proposed, integrating deterministic (energy-efficient) and probabilistic (distributionally robust) insights. Simulation results demonstrate superior performance compared to baselines, achieving over three-fold energy savings.

Publisher

Published On

Authors

Mohamed Badi, Chaouki Ben Issaid, Anis Elgabli, Mehdi Bennis

Related Publications

Explore these studies to deepen your understanding of the subject.

Political Science

Policy failures and energy transitions: the regulatory bricolage for the promotion of renewable energy in Mexico and Chile

M. Natorski and I. Solorio

Psychology

Dopamine release and dopamine-related gene expression in the amygdala are modulated by the gastrin-releasing peptide in opposite directions during stress-enhanced fear learning and extinction

Y. Morishita, I. Fuentes, et al.

Psychology

Dopamine release and dopamine-related gene expression in the amygdala are modulated by the gastrin-releasing peptide in opposite directions during stress-enhanced fear learning and extinction

Y. Morishita, I. Fuentes, et al.

Economics

Unveiling the nexus: exploring the influence of terrorism on energy trade in China and the Belt and Road countries

W. Yang, W. Shi, et al.

Listen, Learn & Level Up

Over 10,000 hours of research content in 25+ fields, available in 12+ languages.

No more digging through PDFs, just hit play and absorb the world's latest research in your language, on your time.

listen to research audio papers with researchbunny