Announcement Detail
Thursday, January 8, 2026
3pm EST
Join via Zoom: https://us06web.zoom.us/j/
USACM UQ Virtual Seminar
The loss is not enough: efficient UQ for misspecified surrogate models with POPS
Speaker
Thomas Swinburne, University of Michigan
Abstract
Useful surrogate models are typically efficient but misspecified, i.e. no one parameter choice is able to exactly reproduce observations. As a result, model parameters are fundamentally uncertain, i.e. have model-form error, as there is no unique “best choice”. In addition, finite capacity models (e.g. polynomials, feature models) are often underparametrized meaning epistemic uncertainties are minimal.
I will discuss recent work[1] which treats the true generalisation error, a misspecification-aware error measure for which the misspecification-blind log likelihood of Bayesian inference is only an (Hoeffding/Jensen) upper bound, closely analogous to the Gibbs-Bogoliubov bound (F
[1] T. Swinburne and D. Perez, Mach. Learn.: Sci. Technol., 6, 015008 (2025)
[2] https://github.com/tomswinburne/POPS-Regression.git
Upcoming webinar speakers:
Monday, February 8: Wei Chen, Northwestern University
Thursday, March 12: Hannah Lu, The University of Texas at Austin
Thursday, April 16: Jinlong Wu, University of Wisconsin-Madison