Neural Networks for Multi-Antenna Channels at High Spectral Efficiency 

Multiple-input and Multiple-output systems (MIMO) consist of multiple transmitting and receiving antennas. They are specific to the field of wireless communication, and they are employed in many of its applications such as 4G, 5G, and Wi-Fi. Neural Networks are models that are trained to recognize specific patterns in large data sets, such that their parameters are adjusted to then mimic the performance of a certain system and predict the output. The problem with MIMO systems is that the complexity of calculation required for detection increase drastically as the number of antennas increases and the complexity of the modulation schemes increases. In our solution to this problem, we propose building an AI decoder that employs a neural network for high speed performance and reliability in detection that is very close to the optimal decoder. This solution is only a computer simulation. Our AI decoder works on signal detection in Additive White Gaussian Noise (AWGN) channels for BPSK and QPSK modulations and for 16 and 64 QAM.