Deep Diffusion Probabilistic Learning for Resource Allocation in Multi-Carrier NOMA Systems

This paper introduces a deep learning framework for resource allocation in multi-carrier non-orthogonal multiple access (NOMA) systems. We use diffusion probabilistic models to address this challenging optimization problem. Our approach transforms resource allocation into a structured prediction task by using channel state information as input features. The diffusion model learns to generate optimal channel-selection matrices. Key contributions include adapting models to wireless communications through a modified loss function that enforces NOMA-specific constraints and a spatial expansion technique for processing wireless resource matrices. Numerical evaluations demonstrate superior performance, with over 90% accuracy in joint channel assignment and power allocation across diverse network configurations. The proposed method significantly outperforms both conventional optimization approaches and traditional deep learning techniques, offering an effective solution for resource management in next-generation wireless networks.