Diffusion Probabilistic Models for Resource Allocation in Multi-Carrier NOMA Systems

This paper presents a machine learning-based approach for resource allocation in multi-carrier non-orthogonal multiple access (NOMA) downlink systems. We propose a diffusion probabilistic model designed to optimize channel assignment and power allocation jointly. Our system model practically accommodates two users per channel, balancing performance and complexity. The diffusion probabilistic model adapts recent advancements in generative models to the resource allocation task, treating it as an image segmentation problem. Our approach incorporates techniques to ensure adherence to the specific structure of NOMA channel assignments. We present numerical results demonstrating the effectiveness of our proposed method in identifying near-optimal channel assignments across various network configurations. The framework significantly outperforms conventional approaches in channel assignment accuracy with sufficient training data. This work advances the integration of advanced machine learning techniques in wireless communications optimization, offering promising solutions for effective resource allocation in NOMA systems.