Deep Learning for massive MIMO: Challenges and Future prospects

The wireless networks today are complex, massive and have dynamic capacity demands. Increase in demand resulted into trouble in managing and monitoring the network components. Thus, intelligent data-driven designs and approaches are required so that the 5th generation (5G) of mobile systems can be reformed for enabling self-organizing capabilities. Thus, in the last decade, mathematical models are designed and adapted among modems. This paper presents a comprehensive outline of the emerging research on deep learning-based models for massive MIMO systems. In most of the work, Deep learning models are used for redesigning the conventional communication system. It may involve channel encoding, decoding, detection, recognition, antenna selection, modulation, etc. It will be interesting to comprehend that replacement of the communication system with a profoundly new architecture such as deep learning based autoencoder, convolutional neural network, etc. These Deep learning-based models show promising performance enhancements with a few limitations and can be efficiently used with massive MIMO.