Deep Learning for the Physical Layer

Lecturer: 
Dr Jakub Hoydis
Date: 
20.2.2019 (All day)

(Place and time: to be defined)

Abstract

In the last decade, deep learning has led to many breakthroughs in various domains, such as computer vision, natural language processing, and speech recognition. Motivated by these successes, researchers all over the world have recently started to investigate applications of this tool to their respective domain of expertise, with communications being one of them. The goal of this tutorial is to provide an introduction to deep learning that will enable the attendees to identify potential applications in their own research field. We give an overview of the very rapidly growing body of literature, explain state-of-the-art neural network architectures and training methods, and go through several promising applications and concepts, such as neural decoding, deep MIMO detection, autoencoders, and information bottleneck. The attendees receive tutorial slides and Jupyter notebooks containing code examples which allows them to quickly get up to speed with this new and exciting field. During the break, we demonstrate the world's first fully neural network-based communications system.

Last updated: 16.1.2019