Clustering-Based Evolutionary Federated Multiobjective Optimization and Learning

2025-04-29

AI Summary

Addressing the multi-objective optimization challenge in federated learning—where communication efficiency, model accuracy, and privacy preservation are difficult to jointly optimize—this paper proposes FedMOEAC. The framework introduces a novel clustering-enhanced evolutionary mechanism to sustain population diversity and is the first to synergistically integrate model quantization, weight sparsification, and a differentially private stochastic gradient descent (DP-SGD) variant into an improved NSGA-II framework for Pareto-optimal joint optimization of all three objectives. Evaluations on MNIST and CIFAR-10 demonstrate a model accuracy of 98.2%, a 45% reduction in communication overhead, a privacy budget ε < 1.0, and a 33% acceleration in convergence compared to the baseline NSGA-II. FedMOEAC establishes a scalable, multi-objective optimization paradigm for efficient and privacy-preserving federated modeling in privacy-sensitive applications.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Search and Optimization: Learning to Search
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Security and Privacy: Security and privacy of machine learning and AI applications
• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing

Abstract

Federated learning enables decentralized model training while preserving data privacy, yet it faces challenges in balancing communication efficiency, model performance, and privacy protection. To address these trade-offs, we formulate FL as a federated multiobjective optimization problem and propose FedMOEAC, a clustering-based evolutionary algorithm that efficiently navigates the Pareto-optimal solution space. Our approach integrates quantization, weight sparsification, and differential privacy to reduce communication overhead while ensuring model robustness and privacy. The clustering mechanism en-hances population diversity, preventing premature convergence and improving optimization efficiency. Experimental results on MNIST and CIFAR-10 demonstrate that FedMOEAC achieves 98.2% accuracy, reduces communication overhead by 45%, and maintains a privacy budget below 1.0, outperforming NSGA-II in convergence speed by 33%. This work provides a scalable and efficient FL framework, ensuring an optimal balance between accuracy, communication efficiency, and privacy in resource-constrained environments.

Problem

• Balancing communication efficiency, model performance, and privacy in federated learning
• Formulating FL as a federated multiobjective optimization problem
• Reducing communication overhead while ensuring model robustness and privacy

Innovation

• Clustering-based evolutionary algorithm for multiobjective optimization
• Integrates quantization, sparsification, and differential privacy
• Enhances diversity and prevents premature convergence