Date: 2025-11-14

Time: 15:30-16:30 (Montreal time)

Location: In person, Burnside 1104

https://mcgill.zoom.us/j/82687773039

Meeting ID: 826 8777 3039

Passcode: None

Abstract:

Deep learning is currently transforming how inverse problems arising in imaging reconstruction are solved. However, it is increasingly well-known that such deep learning-based methods are susceptible to hallucinations. In this talk, I will present a series of theoretical explanations for why hallucinations occur, in both deterministic and statistical estimators. I will conclude by observing that hallucinations can only be avoided by careful design of the forwards operator in tandem with the recovery algorithm, and then provide a theoretical framework for how this can be achieved when solving inverse problems using generative models.

Speaker

Ben Adcock is a Professor of Mathematics at Simon Fraser University. He studied mathematics at the University of Cambridge, where he received his PhD in 2011 as a member of the Numerical Analysis group under the supervision of Arieh Iserles.

He has received numerous awards, including the CAIMS/PIMS Early Career Award (2017), an Alfred P. Sloan Research Fellowship (2015), and a Leslie Fox Prize in Numerical Analysis (2011). Professor Adcock has authored and co-authored three books and monographs. Among them, his book Compressive Imaging: Structure, Sampling, Learning was a finalist for the Association of American Publishers Awards for Professional and Scholarly Excellence (PROSE Awards).

He serves on the Board of Directors of the Society for the Foundations of Computational Mathematics and is a member of the editorial boards of the SIAM Journal on Scientific Computing, the IMA Journal of Numerical Analysis, Numerische Mathematik, the Journal of Approximation Theory, the SIAM Computational Science and Engineering book series, and the CMS/CAIMS Books in Mathematics series (Springer).