Tia Chung (Emory): Latent Twins
Please note this event occurred in the past.
March 13, 2026 11:00 am - 12:00 pm ET
LGRT 1685
Abstract
Abstract:
Latent Twins is a framework that jointly learns low-dimensional representations and the time-evolution operators governing scientific models. This co-adaptation yields compact, interpretable, evolution-aware surrogates that approximate full solution operators, not just one-step predictors.
The approach unifies ideas from reduced-order modeling, paired autoencoders, and operator learning. We will discuss applications ranging from nonlinear ODEs to the shallow-water equations and show how latent-space inference enables prediction from sparse and noisy observations.
Latent Twins bridge representation learning and scientific computing, combining efficiency, interpretability, and operator-level generalization in a single end-to-end framework.
Speaker: Matthias (Tia) Chung is an Associate Professor in the Department of Mathematics at Emory University. His research lies within this cross-disciplinary field of inverse problems, which aims at inferring information from the physical model given observations. Techniques developed in this field are of increasing interest to communities such as system biology, systems engineering, and medical and geophysical imaging, to name a few.