Neural Networks and Operators: Analysis, Algorithms and Applications

各位老师同学好, 以下发布一则学术报告通知, 欢迎各位老师同学参加, 谢谢!

报告题目:Neural Networks and Operators: Analysis, Algorithms and Applications

报告人:何俊材 助理教授 清华大学

报告时间:2026年1月5日 下午16:00-17:00

报告地点:红瓦楼726

报告内容简介:This talk presents recent advances in the theoretical foundations and numerical analysis of deep neural networks and neural operators. We begin with results on approximation and generalization for neural networks, several of which are motivated by classical finite element methods. Building on this perspective, we introduce a unified framework that connects convolutional neural networks with multigrid methods for partial differential equations, known as MgNet. We then highlight recent progress on both the theoretical analysis and algorithmic applications of MgNet, including the first rigorous quantitative approximation results for CNNs with two-dimensional inputs and the development of MgNO, an efficient multigrid-based neural operator framework.

报告人简介:何俊材2014年本科毕业于四川大学,2019年博士毕业于北京大学。2019年至2020年,在宾夕法尼亚州立大学从事博士后研究;2020年至2022年,在得克萨斯大学奥斯汀分校担任R. H. Bing Instructor;2022年至2025年,在沙特阿卜杜拉国王科技大学任研究科学家;2025年2月,加入清华大学丘成桐数学科学中心,任助理教授。他的主要研究方向是科学计算与机器学习,涵盖深度神经网络的理论分析、算法设计与实际应用等方面。

报告邀请人:黄学海