Mathematical Sciences Department Colloquium - Xueyu Zhu, University of Iowa "Efficient Bayesian Physics Informed Neural Networks for Inverse Problems via Ensemble Kalman Inversion" via Zoom: 938 8407 5985
11:00 am to 12:00 pm
Mathematical Sciences Department
Speaker: Xueyu Zhu, University of Iowa
Friday, September 29, 2023
11:00 am - 12:00 pm
Virtual Colloquium, via Zoom: 938 8407 5985
Title: Efficient Bayesian Physics Informed Neural Networks for Inverse Problems via Ensemble Kalman Inversion
Abstract : Bayesian Physics Informed Neural Networks (B-PINNs) have gained significant attention for PDE-based inverse problems. Existing inference approaches are either computationally expensive for high-dimensional posterior inference or provide unsatisfactory uncertainty estimates. In this paper, we present a new efficient inference algorithm for B-PINNs that uses Ensemble Kalman Inversion (EKI). We find that our proposed method can achieve inference results with informative uncertainty estimates comparable to Hamiltonian Monte Carlo (HMC)-based B-PINNs with a much reduced computational cost.
Short bio of the speaker: Professor Xueyu Zhu is an associate professor at the Department of Mathematics, University of Iowa. He is also affiliated with interdisciplinary Ph.D. programs in Applied Mathematical and Computational Sciences and the Iowa Initiative for Artificial Intelligence. His research interests lie in computational mathematics, scientific computing and data science. He received his PhD from Division of Applied Mathematics at Brown University, supervised by Professor Jan S Hesthaven. From 2013 to 2016, He did his postdoc under the supervision of Professor Dongbin Xiu at University of Utah.
Faculty host: Qiao Zhuang