Date: Friday, April 4, 2025
Time: 10:30–12:00 Pacific time
Location: Online or in person at the Earth Sciences Building (ESB) 5104, 2207 Main Mall, University of British Columbia, Vancouver, B.C.
This special event represents a convergence of the CANSSI SSC Seminar on Innovations in Statistics and Data Science and the Constance van Eeden Seminar, an annual event held at the University of British Columbia.
The CANSSI SSC Seminar is a new series co-sponsored by CANSSI and the Statistical Society of Canada (SSC) that brings distinguished researchers in statistical sciences to CANSSI member universities across Canada. The series promotes interactions between leading researchers and statistical sciences faculty members and students, particularly at smaller institutions.
The Constance van Eeden seminar is a yearly event in which graduate students from the University of British Columbia (UBC)’s Department of Statistics vote for their favourite statisticians. The winner is contacted by the organizing committee and invited to give a talk in the department’s seminar. The speaker spends one or two days on campus, and graduate students have the opportunity to have lunch and dinner with them.
To register for online or in-person participation, visit the event web page.
This year’s speaker is Dr. Arnaud Doucet, Professor of Statistics at the University of Oxford and Senior Staff Research Scientist at Google DeepMind. Dr. Doucet’s research interests lie in the development and analysis of efficient computational methods for inference and learning, machine learning, signal processing, and related areas.
His talk is titled “From Diffusion Models to Schrödinger Bridges—When Generative Modeling Meets Optimal Transport.”
Denoising 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.