Deep Learning and Computations of PDE (S. Mishra, ETH Zürich)

Sep 02
02-09-2020 12:30 Uhr bis 13:30 Uhr

Deep Learning and Computations of PDE

Speaker: Prof. Dr. Siddhartha Mishra
Affiliation: ETH Zurich, Switzerland
Zoom: Meeting ID: 920 3154 3453 , Passcode: 571616

Abstract: Deep neural networks are rapidly emerging as a key tool in enhancing computations of PDEs. In this talk, we survey recent developments in my research group on two related themes. First, we use supervised learning to approximate observables for PDEs efficiently with applications to UQ and PDE constrained optimization. Second, we discuss recently introduced physics informed neural networks (PINNs) as an efficient discretization tool for both forward as well as inverse problems for certain classes of PDEs.