Clustered Federated Learning: A Review
Majid Morafah, Mahdi Morafah · Artificial intelligence · 2025
Clustered Federated Learning (CFL) has emerged as a powerful extension of traditional federated learning to address the challenges posed by heterogeneous, non-IID data across distributed clients. This chapter provides a comprehensive review of the state-of-the-art CFL methods, categorizing them into model-based, feature-based, and hybrid approaches. Model-based clustering leverages client model updates to form clusters, while feature-based methods utilize client data characteristics, and hybrid approaches integrate both aspects to achieve robust clustering. The chapter also discusses the evaluation metrics and benchmarks used to assess CFL performance, such as accuracy, personalization, and cluster quality, along with case studies demonstrating CFL’s applicability in diverse domains like healthcare, IoT, and autonomous systems. We identify key challenges in CFL, including scalability, dynamic clustering, and privacy preservation, and propose future research directions to further enhance the effectiveness and scalability of CFL frameworks. Overall, this chapter aims to provide a deep understanding of CFL, highlighting its potential to improve federated learning outcomes in complex, real-world scenarios with non-IID data.