报告题目：Shallow and Deep Neural Network Approximations

时间：4月21日上午10:00-11:30

地点：海宇楼602

报告人：Narayanasami Sukumar教授，University of California, Davis

报告人简介：

Prof. Sukumar has published 105 peer-reviewed articles in major international journals with 17650 citations, leading to an h-index of 53 (Source: Google Scholar). His research foci over the past decade has been on novel discretization methods (such as virtual element methods) on polytopal meshes, smooth maximum entropy approximation schemes, construction of high-order cubature rules over polytopes and curved geometries, and new methods development to solve the Kohn-Sham equations of density functional theory. A recent emphasis is on applying deep learning to solve partial differential equations over complex geometries.

报告摘要：

In scientific machine learning (SciML), physics-informed neural networks have become a powerful and emerging construct to solve systems of ordinary and partial differential equations. Underlying SciML are neural networks, which are nonlinear approximations. In this lecture, I will introduce the essential concepts that underlie the building blocks of neural networks. To solve differential equations with PINNs, the use of shallow and deep neural networks (fully-connected feedforward multilayer perceptron, MLP) will be discussed.

作为纪念钱令希院士110周年诞辰系列学术活动，“令希力智讲坛”是大连理工大学工程力学系为促进高水平学术交流、推动学科前沿发展、深化产学研融合和统筹教育人才科技一体化而设立的高端学术平台，旨在汇聚学术界与工业界的杰出专家学者，共同探索数智时代的基础研究、技术创新与工程应用的深度融合路径，以推动创新链、产业链与人才链的协同共进。