Autoencoder‐based deep learning for massive multiple‐input multiple‐output uplink under high‐power amplifier non‐linearities

In this paper, the authors study compensation of high-power amplifier (HPA) non-linear distortion in multi-user (MU) massive multiple-input multiple-output (MIMO) systems and focus on uplink transmission, where base station (BS) uses a large antenna array. First, present iterative cancellation (NDIC) algorithm-based MMSE approximate message passing (AMP) at receiver level, order to estimate mitigate jointly channel noise. Second, propose novel technique based deep learning. At first introduce multilayer neural network, trained Levenberg–Marquardt algorithm by eliminating HPA non-linearities ‘Pre distortion’ transmitter ‘Post side. Next, developed end-to-end (E2E) learning approach for joint non-coherent Rayleigh fading channel. The basic idea lies use networks (DNNs), auto encoder (AE) unknown channels, DNNs are applied perform several functions modules existing transmission chain. simulation results demonstrate strong potential proposed E2E terms improving link quality symbol error rate (SER) compared other techniques presented work.