End-to-End Auto-Encoder Communications with Interference Suppression

This talk will present an end-to-end communications system that is modeled as an autoencoder for which the transmitter and receiver functionalities are trained as deep neural networks of the encoder and decoder, respectively. Practical design aspects will be discussed regarding training data limitations, embedded implementation constraints, as well as channel and interference (jamming) effects. The benefits of auto-encoder based communications to operate effectively in the presence of unknown and dynamic interference effects will be highlighted to support next-generation communication systems.