Deep Learning for the Degraded Broadcast Channel

TL;DR

This paper demonstrates that deep auto-encoders can effectively learn to communicate over a multiuser degraded broadcast channel, optimizing coding, labeling, and power allocation based on user SNRs.

Contribution

It introduces a deep learning approach to jointly optimize transceiver design, including coding, labeling, and power allocation, for the degraded broadcast channel.

Findings

Auto-encoders learn to communicate using superposition coding.

Neural networks optimize bit labeling and power allocation based on SNR.

Deep learning achieves effective multiuser broadcast communication.

Abstract

Machine learning has shown promising results for communications system problems. We present results on the use of deep auto-encoders in order to learn a transceiver for the multiuser degraded broadcast channel, and see that the auto encoder is able to learn to communicate on this channel using superposition coding. Additionally, the deep neural net is able to determine a bit labeling and optimize the per user power allocation that depends on the per user SNR.