Deep Learning Design for Joint Antenna Selection and Hybrid Beamforming in Massive MIMO

In this paper, we propose a deep-learning-based for joint antenna selection and hybrid beamformer design problem in mmWave massive MIMO systems. In this respect, we treat both problems as a classification problem. We design two convolutional neural networks (CNNs) which accept the input as the channel matrix and it yields the output as the optimum antenna subarray. The selected part of channel matrix is fed to the second CNN which gives the output as the analog and baseband beamformers. We evaluate the performance of the proposed approach through numerical simulations and show that our CNN framework provides significantly better performance as compared to the conventional techniques such as orthogonal matching pursuit.