Beam Allocation in THz-NOMA Networks: A Graph Neural Network Approach

In this paper, we propose a graph neural network (GNN)-based algorithm for beam resource allocation in a terahertz communication system that employs non-orthogonal multiple access (NOMA). The proposed GNN architecture, consisting of graph attention layers and fully connected layers, is designed to map the features of users and beams to optimal power allocation strategies. Compared to conventional methods based on convex relaxations, our algorithm enhances the sum rate performance and computational efficiency. The trained network also adapts to varying numbers of secondary users without retraining, making it suitable for dynamic communication scenarios.