부산대학교 도서관

This paper introduces Quantum Deep Unfolding (QDU), a technique for optimizing power allocation and transmit precoding in multiple-input multiple-output non-orthogonal multiple access (MIMO-NOMA) systems. Solving the optimization problem in such systems poses a significant challenge due to its high computational complexity and non-convex nature, which increases the risk of being stuck at a local minimum. In order to address this issue, QDU leverages an iterative algorithm and analytical derivation to enhance the sum rate performance and training processes by optimizing power allocation and transmit precoding. The proposed approach integrates a Quantum Neural Network (QNN) induced by an iterative deep unfolding algorithm with a learning solution inspired by the training process. At each QDU layer, the iterative optimization involving the Projected Gradient Descent (PGD) operator is unfolded to learn the crucial parameters. The objective of QDU is to maximize the achievable sum rate while simultaneously reducing computational complexity.