Intelligent Earth Seminar: Arnaud Doucet (University of Oxford)

10 March 14:00
Seminar Room 1a, Doctoral Training Centre, 1-4 Keble Rd

Arnaud Doucet received his PhD in Information Engineering from University Paris-Sud in 1997. Even since, he has held faculty positions at  Melbourne University, Cambridge University, the University of British Columbia, the Institute of Statistical Mathematics and Oxford University where he was a statutory professor in the department of Statistics. He is currently a Senior Staff Research Scientist at Google DeepMind. His main research interests are computational statistics, generative modeling and Monte Carlo methods. He was a 2016 Institute of Mathematical Statistics (IMS) Medallion lecturer, was awarded the Guy Silver Medal by the Royal Statistical Society in 2020, the Akaike Memorial award by the Japanese Statistical Society in 2024 and was the Breiman Lecturer at NeurIPS in 2024.

From Diffusion Models to Schrodinger Bridges - Generative Modeling Meets Optimal Transport

Diffusion models have revolutionized generative modeling. Conceptually, these methods define a transport mechanism from a noise distribution to a data distribution. Recent advancements have extended this framework to define transport maps between arbitrary distributions, significantly expanding the potential for unpaired data translation. However, existing methods often fail to approximate optimal transport maps, which are theoretically known to possess advantageous properties. In this talk, we will show how one can modify current methodologies to compute Schrödinger bridges, an entropy-regularized variant of dynamic optimal transport. We will demonstrate this methodology on a variety of unpaired data translation tasks.