Graph Neural Networks for Wireless Networks: Graph Representation, Architecture and Evaluation
Lu Yang, Yuhang Li, Ruichen Zhang, Wei Chen, Bo Ai, Dusit Niyato
Abstract
Graph neural networks (GNNs) have been regarded as the basic model to facilitate deep learning (DL) for revolutionizing resource allocation in wireless networks. GNN-based models are shown to be able to learn the structural information about graphs representing the wireless networks to adapt to the time-varying channel state information and dynamics of network topology. This article aims to provide a comprehensive overview for applying GNNs to optimize wireless networks via answering three fundamental questions, that is, how to input the system parameters of wireless networks into GNNs, how to improve the expressive performance of GNNs, and how to evaluate GNNs. Particularly, two graph representations are given to transform wireless network parameters into graph-structured data. Then, we focus on the architecture design of the GNN-based models by introducing the basic message passing, as well as model improvement methods, including a multi-head attention mechanism and residual structure. Finally, we give task-oriented evaluation metrics for DL-enabled wireless resource allocation schemes. We also highlight certain challenges and potential research directions for the application of GNNs in wireless networks.