Federated Learning within Global Energy Budget over Heterogeneous Edge Accelerators

2025-06-12

AI Summary

To address device and data heterogeneity, global energy constraints, and multi-objective co-optimization in federated learning (FL) over edge heterogeneous accelerators, this paper proposes the first client selection framework incorporating explicit energy budget constraints. We formulate a novel bi-level integer linear programming (ILP) model that jointly optimizes model accuracy, training latency, and energy consumption. To quantify client contribution under non-IID data, we introduce an approximation of the Shapley value; further, we design an energy–time joint prediction model to enable dynamic, context-aware client selection. Evaluated under realistic edge configurations and non-IID data distributions, our method achieves a 15% improvement in model accuracy and a 48% reduction in training time compared to state-of-the-art approaches, while significantly enhancing the energy-efficiency–performance trade-off—thereby advancing sustainable AI deployment at the edge.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Federated Learning (FL) enables collaborative model training across distributed clients while preserving data privacy. However, optimizing both energy efficiency and model accuracy remains a challenge, given device and data heterogeneity. Further, sustainable AI through a global energy budget for FL has not been explored. We propose a novel optimization problem for client selection in FL that maximizes the model accuracy within an overall energy limit and reduces training time. We solve this with a unique bi-level ILP formulation that leverages approximate Shapley values and energy-time prediction models to efficiently solve this. Our FedJoule framework achieves superior training accuracies compared to SOTA and simple baselines for diverse energy budgets, non-IID distributions, and realistic experiment configurations, performing 15% and 48% better on accuracy and time, respectively. The results highlight the effectiveness of our method in achieving a viable trade-off between energy usage and performance in FL environments.

Problem

• Optimize energy efficiency and model accuracy in Federated Learning
• Address device and data heterogeneity in FL training
• Maximize FL model accuracy within global energy budget

Innovation

• Bi-level ILP formulation for client selection
• Approximate Shapley values for optimization
• Energy-time prediction models for efficiency