Introduction to Deep Learning Possibilities in Communication Systems

Abstract

The considerable requests imposed on modern communication systems have inspired re-evaluation of the existing well-defined and elaborated communication theory. It has become debatable whether the current system implementations have become obsolete or may be able to cope with the challenges if combined with new technologies. The idea of developing entirely new approaches has also risen. Recent advances in machine learning and especially deep learning techniques indicate possible new research directions. This paper states the reasons behind the introduction of deep learning in communications. The paper provides the separation of the existing research procedures. Special focus is put on analyzing different areas of appliance and defining their advantages and disadvantages, including the formation of end-to-end communication systems as autoencoders and neural network contribution in signal detection and modulation classification. The results indicate strong usage potential of deep learning in communications in not already optimized areas.