BIMSA >
Seminar on Control Theory and Nonlinear Filtering
When and why PINNs fail to train from the viewpoint of neural tangent kernel
When and why PINNs fail to train from the viewpoint of neural tangent kernel
Organizer
Speaker
Time
Friday, November 17, 2023 10:00 PM - 10:30 PM
Venue
Online
Abstract
In this talk, we will share several literatures about the explanation of training dynamics of Physics-informed neural network. Especially, as a uniform framework, neural tangent kernel demonstrates the dynamical behavior when width of NN goes to infinity in the single layer setting. NTK illustrates training one-layer NN with infinite width is the same as a kernel regression. As a consequence, NTK will exhibit the training weights of different parts in loss function which will provide the guidance for adaptive algorithm of revised PINN.
Speaker Intro
Jiao Xiaopei graduated with a bachelor's degree from the Zhi Yuan College of Shanghai Jiao Tong University (Physics Department) in 2017 and obtained his PhD from the Department of Mathematical Sciences at Tsinghua University in 2022, under the guidance of Professor Stephen Shing-Toung Yau (IEEE Fellow, former tenured professor at the University of Illinois at Chicago). He has conducted postdoctoral research at the Beijing Institute of Mathematica Science and Application and at the University of Twente in the Netherlands (under the guidance of Professor Johannes Schmidt-Hieber, Fellow of the Institute of Mathematical Statistics). His current research interests include control theory, numerical partial differential equations, and bioinformatics.