Biosketch
Christopher began his PhD at the Chair of Mathematical Foundations of Artificial Intelligence in early 2024 and soon after joined relAI. His research is dedicated to utilizing neural networks to learn and describe physical laws and systems.
Christopher holds a Bachelor’s degree in Industrial Engineering and a Master’s degree in Economathematics from Karlsruhe Institute of Technology. During his studies, he gained valuable experience as a research assistant, focusing on uncertainty quantification and the use of neural networks for weather system forecasting.
relAI Research
Uncertainty quantification for neural operators
Reliable uncertainty quantification is essential for deploying machine learning in safety-critical scientific applications, yet principled approaches remain scarce — particularly in structured, high-dimensional output spaces such as spatiotemporal systems. Our research develops new methodologies for probabilistic prediction and uncertainty quantification centered on proper scoring rules. We introduce the Probabilistic Neural Operator, a framework that extends neural operator architectures to learn full predictive distributions over function spaces by training with proper scoring rules. Building on this foundation, we develop a unified framework for decomposing predictive uncertainty into aleatoric and epistemic components, offering principled design guidelines for task-specific uncertainty measures. We further explore diffusion-based generative models as flexible tools for probabilistic regression. Applications span weather forecasting, chaotic dynamical systems, and spatial extremes.
Publications
https://openreview.net/forum?id=gangoPXSRw
https://doi.org/10.1175/AIES-D-24-0049.1