Mathematical Theory of Machine Learning

The project aims to contribute to the development of theoretical frameworks that explain the emergent properties of modern machine learning methods involving deep artifical neural networks.

The ongoing AI revolution was sparked in the early 2010s by large and deep Artificial Neural Networks. Their development has been driven largely by empirical experimentation, rather than by a fundamental theoretical understanding of their emergent properties. This research project aims to contribute to the development of a comprehensive theory of modern neural networks, with a focus on scaling limits and novel training algorithms.

Mitarbeiter

Prof. Dr. David Belius Principal investigator

Dr. Volodymyr Fomichov Collaborator

Georgios Vlassis Collaborator