Workshop on Mathematics of Deep Learning 2019

Deep Learning has evolved into one of the hot topics in industry and science with a wide range of applications related to the processing and interpretation of large amounts of data. This includes recommendation and chat systems, image and language recognition, classification, identification, knowledge discovery, simulation of complex physical phenomena, autonomous systems, and many other areas.

While the success of recent neural network architectures has been breathtaking, the mathematical understanding of these architectures is still in its infancy. However, mathematical disciplines have recently began to embrace this challenge and tremendous progress has already been made, for instance regarding the analysis of the expressive power, approximation properties, complexities and the stochastic training. With such insights, one can hope to further the development of more efficient algorithms, improved architectures and the understanding of the reliability of these modern methods.

The aim of the workshop is to bring together leading researchers with a focus on the mathematical analysis and interpretation of current Machine Learning approaches. While Deep Learning with Artificial Neural Networks seems to be the most popular research area, alternative methods and representations will also be presented.

Here you find the UPDATES (December 4) of the schedule and of the detailed program.

The updated booklet (December 4) (including abstracts and list of participants) can be downloaded HERE. It is only available online.

Invited Speakers

  • Helmut Bölcskei (ETH Zurich)
  • Alexandra Carpentier (Otto von Guericke University Magdeburg)
  • Philipp Grohs (University of Vienna)
  • Arnulf Jentzen (University of Münster)
  • Gitta Kutyniok (TU Berlin)
  • Anthony Nouy (EC Nantes)
  • Ivan Oseledets (Skoltech, Moscow)
  • Philipp Petersen (University of Oxford)
  • Christoph Schwab (ETH Zurich)
  • Ingo Steinwart (University of Stuttgart)
  • Edwin Miles Stoudenmire (Flatiron Institute, New York)
  • Dmitry Yarotsky (Skoltech, Moscow)


  • Martin Eigel (WIAS Berlin)
  • Peter Friz (TU Berlin/WIAS Berlin)
  • Reinhold Schneider (TU Berlin)
  • Volodia Spokoiny (HU Berlin/WIAS Berlin)


We gratefully acknowledge support by Berlin Mathematics Research Center MATH+.


Contact and further information

Phone: +49 30 20372-555