LMS offers course “Physical Models meet Deep Learning” at Ferienakademie 2022

Symbolic picture for the article. The link opens the image in a large view.

In the past decade, deep learning (DL) algorithms have gained increasing popularity in many areas of signal processing and communications and managed to solve tasks which were previously considered infeasible. However, classical model-based approaches such as nonnegative matrix factorization, the Kalman filter or the Viterbi algorithm are still of great relevance as they do not rely on large training datasets and yield well-interpretable results.


In the course “Physical Models meet Deep Learning” at Ferienakademie 2022, we will investigate recent research on the synthesis of DL techniques and model-based algorithms. For example, by carefully replacing parts of model-based algorithms with artificial neural networks, the algorithms’ performance is increased while only requiring small amounts of training data and ensuring an explainable behavior at every step.


The two-week course will be held by Profs. Walter Kellermann (Audio Signal Processing, FAU), Gerhard Kramer (Information and Communication Theory, Technical University Munich) and Stephan ten Brink (Signal Processing and Channel Coding, University Stuttgart). Further information about the Ferienakademie as well as application details (deadline: May 10, 2022) are available on http://www.ferienakademie.de/. For questions regarding the course, please contact Johannes Zeitler (johannes.zeitler@fau.de). We look forward to seeing you in Sarntal!