Unsupervised Learning-Based Resource Allocation for Cognitive Radio Networks

Abstract

Cognitive radio technology allows secondary users (SUs) to opportunistically access licensed spectrum to improve the spectral efficiency of communication systems. In this paper, by utilizing deep neural networks (DNNs), we study the resource allocation of the SUs in cognitive radio networks (CRN) and propose a scheme based on unsupervised learning to maximize the sum rate of the SUs. The proposed scheme ensures that the interference caused to primary users (PUs) does not exceed a predefined threshold. We also discuss the quality of service (QoS) requirements of the SUs. The numerical simulation results show that the proposed scheme achieves a higher sum rate with low computation time.