Join us for a talk by Kyunghyun Cho, Glen de Vries Professor of Health Statistics and Professor of Computer Science and Data Science at New York University. This talk is part of the Kempner Seminar Series, a research-level seminar series that covers topics related to the basis of intelligence in natural and artificial systems.
In this talk, I will present a series of preliminary studies that demonstrate the potential of using deep learning to derive statistical and causal estimators that were once and currently thought out of reach for us to design ourselves, including targeted cause discovery, blackbox causal inference, mutual information estimation by supervised training and trajectory-aware learned acquisition functions. These studies are motivated from the observation that (scientific) discovery problems often call for novel algorithms that we are not aware of or having challenging time designing.