DM3L Seminar: Jonas Peters
Causality, Hidden Confounding, and Robustness
Talk by Prof. Dr. Jonas Peters, ETH Zurich
Date: 30.10.25 Time: 12.00 - 13.15 Room: Y27H12
When learning predictive models on real world data, we usually expect
them to work well on test data that follow the same distribution as the
training data. Often, however, such an assumption is unrealistic or does
not fit well to the scientific question at hand. Instead, we would like
to apply the fitted model to different locations or times or we want to
know how the system reacts under interventions. In this talk, we show
how these two problems (distributional robustness and causality) are
connected via the concept of hidden confounding. We also discuss a few
ideas on how to tackle such problems. The talk hopefully serves as an
inspiration to think about these interesting, general, and important
problems.