DARENet: Data Arrangement Neural Network for Eigenvector-Based CSI Feedback

In 5G communications, the precision of Channel State Information (CSI) feedback is vital, and the massive Multiple-Input Multiple-Output (MIMO) systems rely heavily on this for optimal performance. While eigenvector-based methods using Deep Learning (DL) have innovated CSI feedback mechanisms, they do not fully exploit the intrinsic correlations within CSI that are instrumental for feedback optimization. To bridge this gap, we propose the Data ARrangEment Neural Network (DARENet), a Convolutional Neural Network (CNN) based framework that effectively utilizes the inherent correlations present in CSI eigenvectors with cross-polarized antennas. DARENet’s capabilities are validated through rigorous testing on 5 public datasets, where it consistently outperforms the established EVCsiNet and PolarDenseNet in terms of recovery performance, computational efficiency, and model complexity.