Federated-Learning-Enabled Intelligent Fog Radio Access Networks: Fundamental Theory, Key Techniques, and Future Trends
Zhongyuan Zhao, Chenyuan Feng, Howard H. Yang, Xueting Luo
Abstract
The rise of big data and AI boosts the development of future wireless networks. However, due to the high cost of data offloading and model training, it is challenging to implement network intelligence based on the existing centralized learning strategies, especially at the edge of networks. To provide a feasible solution, a paradigm of federated learning- enabled intelligent F-RANs is proposed, which can take full advantage of fog computing and AI. The fundamental theory with respect to the accuracy loss correction and the model compression is studied, which can provide some insights into the design of federated learning in F-RANs. To support the implementation of federated learning, some key techniques are introduced to fully integrate the communication, computation, and storage capability of F-RANs. Moreover, future trends of federated learning-enabled intelligent F-RANs, such as potential applications and open issues, are discussed.