The seminar was co-organized by Philipp Grohs, who gave a short series of lectures covering the basics of statistical learning theory and deep learning, and advanced topics such as the mathematical analysis of the approximation properties of deep neural networks, or novel approaches to solving high-dimensional PDEs using deep neural networks which in certain cases can provably break the curse of dimensionality.

The week in Oberwolfach was a great opportunity to discuss emerging ideas and obstacles in the mathematical analysis of deep learning and to get fully immersed in this exciting new research area. It was an ideal starting point for the first phase of our project which now begins at the Unviersity of Vienna.