A review of federated learning: architectures, challenges, and targeted solutions

Abstract

Federated Learning (FL) has emerged as a promising paradigm for distributed intelligence, enabling collaborative model training across multiple clients without transferring raw data to a central server. By preserving data locality, FL addresses fundamental limitations of conventional centralized machine learning, particularly with respect to privacy protection, security exposure, communication overhead, and the management of large-scale heterogeneous data. With the rapid expansion of FL into diverse application domains, ensuring the trustworthiness, scalability, and performance of federated systems has become increasingly important. This paper presents a review of federated learning through a structured taxonomy that covers its core architectural paradigms, major learning types, model training approaches, and aggregation mechanisms. Moreover, it analyzes the principal challenges confronting FL, including privacy and security risks, statistical and system heterogeneity, communication constraints, and global model divergence. In response to these issues, the review further discusses targeted solutions designed to address each challenge and improve the efficiency and reliability of federated frameworks. By combining taxonomy-driven analysis with challenge-oriented solution mapping, this study provides an insightful reference for advancing both the theoretical understanding and practical deployment of federated learning systems.