A Survey about Deep Learning for Constellation Design in Communications

The performance of communication systems depends on the choice of constellations, designed in an end-to-end manner. In case of a mathematical intractability, either because of complexity or even lack of channel model only sub-optimal solutions can be provided with an analytical approach. We present end-to-end learning, a recent technique in communications to learn optimal transmitter and receiver architectures based on deep neural networks (DNNs) architectures. We discuss the cases in which this technique has been used to design constellations, where a mathematical analysis is repressed due to the channel model intractability.