Our research is primarily focused on deep learning, with a core objective being the advancement of the foundations of deep learning. Despite the groundbreaking capabilities of modern deep neural networks, our theoretical understanding is still limited. However, we are convinced that a thorough theoretical understanding is needed for the trustworthy application of existing learning algorithms and the development of novel robust and efficient learning approaches.
Another core area of our research is the intersection of machine learning and security. Here, we are interested in topics like uncertainty quantification, robustness against adversarial attacks, and defense strategies against the threats induced by artificial media generated by machine learning systems.