Deep Models and Optimization

Investigating the interplay between optimizer and architecture in Deep Learning, and new networks for long-range reasoning.

The purpose of our research is to design new optimizers and neural networks to accelerate science and technology with Deep Learning. Our approach is theoretical, with a focus on optimization theory as a tool for dissecting the challenging dynamics of modern foundation models. By developing new technologies grounded in theoretical knowledge, we envision a future in which scientists and engineers, regardless of resource constraints, can leverage powerful and reliable deep learning solutions to help make the world a better place.

Teaching at the University of Tübingen: Nonconvex Optimization for Deep Learning (Winter Semester 24/25 / 25/26), Details here.

New PhD students: We are not hiring new PhD students at the moment. If you, however, know optimization very well... drop us an email!