Federated Learning With Non-IID Data in Wireless Networks

Federated learning provides a promising paradigm to enable network edge intelligence in the future sixth generation (6G) systems. However, due to the high dynamics of wireless circumstances and user behavior, the collected training data is non-independent and identically distributed (non-IID), which causes severe performance degradation of federated learning. To solve this problem, federated learning with non-IID data in wireless networks is studied in this paper. Firstly, based on the derived upper bound of expected weight divergence, a federated averaging scheme is proposed to reduce the distribution divergence of non-IID data. Secondly, to further harmonize the distribution divergence, data sharing is associated with federated learning in wireless networks, and a joint optimization algorithm is designed to keep a sophisticated balance between the model accuracy and the cost. Finally, the simulation results based on a common-used image data set are provided to evaluate the performance of our proposed schemes, which can achieve significant performance gains with a small price of latency and energy consumption.

I. Introduction

To support emerging intelligent services and applications, artificial intelligence (AI) has been considered as a key technique in the future sixth generation (6G) systems [1]. By employing AI-enabled network management and operation schemes, the communication, computation, and storage capability of networks can be fully integrated in an efficient way [2]. Moreover, to support various AI-enabled applications and services, a paradigm of deploying machine learning at network edge devices has been proposed in [3]. As introduced in [4], the application scenarios of 6G systems are extended, which have to satisfy diverse quality-of-service (QoS) requirements with respect to throughput, latency, and connections. However, the existing network management strategies, such as the conventional optimization theory-based methods, are computation-intensive and cause long latency. Therefore, it is challenging to adaptively satisfy the QoS requirements due to the high dynamics of wireless networks. The implementations of AI-enabled schemes can provide feasible solutions to this problem. In [5] and [6], AI-enabled core network architecture has been designed, which integrates deep learning techniques with network orchestration and traffic control to improve user experience. In [7], an intelligent paradigm of information-centric network has been proposed to fully explore the potential of edge caching. In [8] and [9], the design of intelligent 6G networks has been discussed to build a bridge between deep learning and cloud/edge computing-based networks.