Quantum neural network empowered spectral efficiency analysis in IRS-assisted communication systems

Abstract

The rapid increase in data consumption by mobile users has created a demand for innovative architectures to address this growing need. Concurrently, classical neural network-based intelligent reflecting surfaces (IRS) have gained traction in 6G networks due to their capacity to enhance spectral efficiency (SE). However, relying solely on classical computing for SE calculations imposes limitations on computational rates, stemming from the exponential increase in complexity. In response to this challenge, this paper proposes the use of quantum convolutional neural networks (QCNNs) to improve the SE of communication systems facilitated by IRS configurations. To optimize IRS phase shifts, we first apply singular value decomposition to the received IRS-optimized data. The resulting information is then incorporated into the quantum data and subsequently processed by the parameterized quantum circuit. A dense layer is introduced after the proposed model to boost training performance. The proposed QCNN model demonstrates a significant enhancement in SE while maintaining lower complexity, leveraging the capabilities of qubits in the current era of quantum computing.