ELLIS Distinguished Lecture
Stefan Feuerriegel: Causal ML for learning treatment effects over time
Time: February 16, 2024, 1:00pm (12:00pm CET)
Venue: Lecture hall TU2, Maarintie 8, Aalto University / Zoom
Abstract:
Treatment effect estimation in continuous time is crucial for personalized medicine. For example, in cancer therapy, a physician may base the decisions of applying chemotherapy on whether or not the expected health trajectory will improve after treatment.
In the first part of the talk, we will introduce the foundation of causal ML, the underlying assumptions, and provide practical applications. This is designed as quick primer for those who have not (yet) worked in causal ML.
In the second part, we develop a novel Causal Transformer for estimating counterfactual outcomes over time. Our Causal Transformer addresses limitations of state-of-the-art methods, which build upon simple long short-term memory (LSTM) networks. As a result, our Causal Transformer is designed at inferences for complex, long-range dependencies in patient trajectories.
In the third part, we give a short outlook of how to extend causal ML for probabilistic inferences, so that we can obtain uncertainty estimates. Specifically, we propose a novel Bayesian neural controlled differential equation (BNCDE) for treatment effect estimation in continuous time.
Bio:
Stefan Feuerriegel is a full professor at LMU Munich School of Management at LMU Munich, where he heads the Institute of AI in Management. Previously, he was an assistant professor at ETH Zurich. In his research, Stefan develops, implements, and evaluates Artificial Intelligence technologies to improve decision-making. His group is particularly interested in understanding the effects of treatments, so that treatment decisions can be improved ("causal ML"). His team also brings the causal ML methods into practice with companies such as Booking.com and ABB Hitachi and medical professionals. Stefan's team publishes regularly in NeurIPS, ICML, ICLR but also general science outlets like Nature Communications and medical outlets like Lancet Digital Health or NEJM AI.