Speaker: Cheng Jin, School of Mathematical Sciences, Fudan University & Shanghai Contemporary Applied Math. Key Lab.
Prof. Cheng Jin is a professor at the School of Mathematical Sciences, Fudan University, and the president of the Shanghai Society of Industrial and Applied Mathematics. He is also a Fellow of the Institute of Physics in the UK and an Executive Committee member of the Eurasian Association of Inverse Problems, among other roles. He has previously served as the Vice President of the Chinese Mathematical Society, a Panel member in the Division of Mathematics and Physics of the National Natural Science Foundation of China, a Panel member for the National Science Foundation (NSF) of the United States, and an editorial board member of multiple internationally renowned journals. He has published over 120 papers in academic journals. In 2019, he received the First Prize of the Shanghai Natural Science Award, the Second Prize of the Shanghai Natural Science Award in 2020, and the First Prize of Shanghai Teaching Achievement Award in 2022. He has made significant progress in theoretical analysis of inverse problems for partial differential equations and efficient inversion algorithms for general inverse problems. In terms of applications, he has effectively collaborated with domestic and foreign enterprises such as Nippon Steel and Huawei, achieving outstanding results and earning praise from the industry.
Time:
Monday, March 9th, 2026,15:00-16:00
Organizer
Liao Qifeng
Venue
Room 408, South Building of SCA
Abstract
The rapid development of Learning theory has opened up a new horizon for scientific research. How to apply the idea of learning in scientific computing and propose new ideas for some difficult problems is currently a research direction of great concern. In this report, we mainly introduce some recent related research results of our team: a method for numerical solution of differential equations based on machine learning. In view of the fact that current numerical solution methods (FEM, FD, BEM etc.) do not consider existing exact solutions (fundamental solutions, series solutions etc.), information from actual engineering measurements, and related results that have already been calculated, we propose a numerical solution method based on learning theory, and provide the relevant theoretical framework and algorithm. The results of numerical simulation show that our method has good effects on high-wavenumber problems.