FairEnergy: Contribution-Based Fairness meets Energy Efficiency in Federated Learning

TL;DR

This work tackles the joint challenge of energy efficiency and fair participation in federated learning over wireless edge networks. It introduces FairEnergy, a fairness-aware framework that uses a contribution-based score $s_i^r(\gamma_i^r)$ and a long-term EMA-based participation metric $q_i^r$ to guide joint optimization of client selection, compression level, and bandwidth allocation. The optimization is formulated as a mixed-integer non-convex problem and solved via relaxation to $[0,1]$ with Lagrangian dual decomposition; per-device subproblems are solved by discretizing $\gamma_i^r$ and applying Golden Section Search for $B_i^r$, with dual variables updated by subgradient ascent. Experiments on non-IID data show that FairEnergy achieves higher model accuracy while reducing total communication energy by up to 71%–79% compared with baselines, highlighting the practical impact of coupling contribution-aware selection with long-term fairness in energy-constrained FL.

Abstract

Federated learning (FL) enables collaborative model training across distributed devices while preserving data privacy. However, balancing energy efficiency and fair participation while ensuring high model accuracy remains challenging in wireless edge systems due to heterogeneous resources, unequal client contributions, and limited communication capacity. To address these challenges, we propose FairEnergy, a fairness-aware energy minimization framework that integrates a contribution score capturing both the magnitude of updates and their compression ratio into the joint optimization of device selection, bandwidth allocation, and compression level. The resulting mixed-integer non-convex problem is solved by relaxing binary selection variables and applying Lagrangian decomposition to handle global bandwidth coupling, followed by per-device subproblem optimization. Experiments on non-IID data show that FairEnergy achieves higher accuracy while reducing energy consumption by up to 79\% compared to baseline strategies.

Figures (3)

• Figure 1: Test accuracy per round.
• Figure 2: Energy consumption per round.
• Figure 3: Total energy needed to achieve a target accuracy.