Kristian Warnholz

Supervisors: Prof. Gregor Kasieczka (UHH), Prof. Sarah Heim (DESY, UHH),  Prof. Timo Gerkmann (UHH)

Kristian studied Physics (B.Sc. and M.Sc.) at the University of Hamburg, focusing on machine learning in the field of particle physics in both his theses. He wrote his Master’s thesis on “Extracting the Statistical Significance for Anomaly Detection Methods”.

This project aims at developing and applying uncertainty-aware generative models, combining approaches from the fields of anomaly detection in particle physics and audio enhancement in signal processing. While generative models are used extensively in both fields these days, a common drawback is the lack of uncertainty quantification and robustness of model outputs. Especially in a scientific context, this poses the question of how trustworthy these outputs are and underscores the need for well-calibrated uncertainty estimates.