Published: Feb 25, 2024 by Daning Huang
The SIAM UQ24 occurs in Trieste, Italy. While we could not be physically present there, we contribute a talk on physics-infused reduced-order modeling as part of a minisymposium: “Recent Advances in Surrogate Modeling for UQ of Large-Scale Simulations”.
The central idea is that, conventional surrogates for UQ of dynamical systems only support a limited number of parameters, while we aim to leverage the PIROM methodology to build surrogates that are of infinite-dimensional parameters by directly encoding known dynamics into the model. We connect this idea to the well-known Mori-Zwanzig formalism, and present two algorithms for learning PIROMs from data, that build on our lab member Carlos’ previous work.
The slides are available here.