Probabilistic Numerics — moving BayesOpt expertise to the inner loop by Philipp Hennig

A Google TechTalk, presented by Philipp Hennig, 2022/02/08 ABSTRACT: BayesOpt Speaker Series. Bayesian Optimization experts are Gaussian process experts. And there is much more to do for Gaussian inference in the algorithmic space beyond outer-loop optimization. Using simulation — the solution of differential equations — as an example, I will argue that the advent of machine learning should put probabilistic functionality and theory at the center of numerical computations. This includes decidedly low-level functionality, a chance for BayesOpt experts to move their skills from the outer to the inner loops of computation. About the speaker: Philipp Hennig holds the Chair for the Methods of Machine Learning at the University of Tübingen, and is an adjunct senior research scientist at the Max Planck Institute for Intelligent Systems. Since his PhD with David MacKay in Cambridge, he has been interested in the relationship between computation and inference. Hennig is the deputy speaker of the Cyber Valley initiative of the German State of Baden-Württemberg; an ELLIS fellow, and co-director of the ELLIS Program for Theory, Algorithms and Computation of Learning Machines; as well core faculty of the Tuebingen AI center. His research has been supported by Emmy Noether, Max Planck, and ERC fellowships / grants.