Developing reduced-order PDEs with machine learning-based closure models

Due to unforeseen reasons, the seminar has to be postponed at short notice to 3:30pm.

CASUS Institute Seminar, Jiequn Han, PhD, Center for Computational Mathematics, Flatiron Institute, New York City, USA

Jiequn received his PhD in applied mathematics from Princeton University and continued his academic career as a lecturer in mathematics at the Department of Mathematics. He then became a Flatiron Research Fellow at the Flatiron Institute’s Center for Computational Mathematics. His current research interests focus primarily on solving high-dimensional partial differential equations and machine learning-based multiscale modeling.

Abstract of the talk // This talk explores the promising role of machine learning (ML) in developing reduced-order partial differential equations for multiscale modeling, also known as the closure problem. Focusing on two critical principles – respecting physical symmetry and data exploration – we demonstrate the success of ML-based closure models in turbulence modeling using Reynolds-averaged Navier-Stokes equations and in constructing hydrodynamic models for kinetic equations. The synergy between physical models and ML leads to more accurate, efficient, and generalizable reduced models. In conclusion, Jiequn discusses a few important future research directions.

The seminar will be held exclusively online via Zoom Inc., people interested in the topic have the chance to also join the talk remotely. Please ask for the login details via contact@casus.science.