Colloquium
DATE	2022-04-14　16:10-17:00
PLACE	數學系館 1F3174教室
SPEAKER	賴明治 教授 (國立陽明交通大學理學院院長)（國立陽明交通大學應用數學系）
TITLE	Solving PDEs Using Machine Learning: How and Why? 線上專題演講
ABSTRACT	因疫情改為線上專題演講 In this talk, we will present how to use machine learning methods to solve partial differential equations, especially when PDEs become difficult to handle with traditional numerical methods (such as finite difference, finite element or spectral methods). We start by introducing some fundamental mathematical backgrounds for neural network representations. Then we introduce our developments for machine learning methodologies by considering elliptic interface problems where the solutions are usually not smooth. We have successfully developed two different types of completely shallow neural network to solve above elliptic PDEs. To the best of our knowledge, this is the first work in literature that uses only one hidden layer with moderate number of neurons to solve the problems even in higher dimensions. 會議鏈結: https://nckucc.webex.com/nckucc/j.php?MTID=m1c055bfb317f185a3bedd063fb2651ec 會議號: 2516 078 9817