Supervised Deep Learning for MIMO Precoding

In this paper, we aim to design an end-to-end deep learning architecture for a broadcast MIMO system with precoding at the transmitter. The objective is to transmit interference-free data streams to multiple users over a wireless channel. We propose end-to-end learning of communication systems modelled as a Deep autoencoder network with a novel cost function to achieve this goal. This architecture enables optimization of the transmitter and receiver network weights jointly over a wireless channel. We also introduce a way to precode the transmitter embeddings before transmission. An end-to-end training of the autoencoder framework of transmitter-receiver pairs is employed while training the proposed transmit-precoded MIMO system model. Several numerical evaluations over Rayleigh block-fading (RBF) channels with slow fading are presented to prove this approach. Specific training methods are suggested to improve performance over RBF channels in this paper.

I. Introduction

In the last few years, Machine learning (ML) and Deep learning [1] has gained significant attraction on solving problems in various domains where theory would not explain the correlation between data. Communication has relied on the classical state of the art algorithms to encode and decode the messages over the decades. However, there has been a growing interest in the application of ML in this field too [2] [3]. With the advent of more powerful hardware, deep learning methods are resurfacing as a viable solution since these computation-intensive algorithms can be deployed for real-time applications.