Biosketch
Sarah is a PhD student in Mathias Drton’s group at the TUM Chair of mathematical statistics. She obtained both her Bachelor’s and Master’s degree in Mathematics at TUM. During her studies, she spent a semester at the Université de Montréal in Canada. As a working student at Fraunhofer IKS she was able to gain experience in the field of trustworthy AI in healthcare. Sarah’s research focuses on causality, graphical models, and, in particular, on how to model causal systems with dynamic processes. She joined relAI in May 2024.
relAI Research
Modeling causal systems with dynamic processes
Many causal structure learning and effect estimation methods require the underlying data generating process to be acyclic, prohibiting feedback loops among the variables. This highly restrictive assumption fails to hold in many real-world applications – from the analysis of gene expression data to economic models. Consequently, the quest for alternative methodologies becomes imperative to ensure robustness, transparency, and reliability of causal modeling and inference in complex systems. In my research, I work on novel frameworks to model high-dimensional stochastic systems. By considering equilibria of underlying dynamic processes like diffusions, these models are designed to accommodate cyclic structures in data. We aim to establish their theoretical properties as well as to develop algorithmic frameworks for dependable structure learning in the presence of feedback loops.