Deep learning-based compressive sampling optimization in massive MIMO systems

In this paper, we develop a deep learning framework to optimize the compressive sampling matrix in a massive multiple-input multiple-output (MIMO) system. The optimized compressive sampling matrix is utilized to project high-dimensional data received at the massive MIMO system into a lower-dimensional space so that the directions of arrival and other signal parameters can be efficiently obtained with a reduced hardware complexity. The proposed deep learning approach for optimizing the compressive measurement matrix increases its robustness and generalizability.