Neural network based optimization of transmit beamforming and RIS coefficients using channel covariances in MISO downlink

Abstract

We propose an unsupervised beamforming neural network (BNN) and a supervised reconfigurable intelligent surface (RIS) convolutional neural network (CNN) to optimize transmit beamforming and RIS coefficients of multi-input single-output (MISO) downlink with RIS assistance. To avoid frequent beam updates, the proposed BNN and RIS CNN are based on slow-changing channel covariances and are different from most other neural networks that utilize channel instances. Numerical simulations show that for a small or moderate signal-to-noise ratio (SNR), the proposed BNN with RIS CNN can achieve a sum rate close to that of a system with optimal beams and RIS coefficients. Furthermore, the proposed scheme significantly reduces the computation time.