Xinliang Liu, Doctor, King Abdullah University of Science and Technology & Shenzhen University

Chensong Zhang, Professor

Multigrid Meets Neural Operator: Efficient Learning for Multiscale and High Frequency Problems

14:00-15:00 July 5(Friday), N602

Neural operator methods provide an innovative approach for solving or learning the complex mappings from parameters to solutions in intricate physical systems. This talk will cover the foundational aspects of neural operators, including theo-retical frameworks and algorithmic developments, with a focus on prominent neural operator architectures. I will also present our recent work on applying the neural operator method to multiscale and high frequency problems. To address the challenges posed by the multiscale and global nature of these problems, we have developed a neural operator (MgNO) with a multigrid structure. We demonstrate the efficiency and accuracy of our method, achieving consistently state-of-the-art performance on multiscale and high frequency partial differen-tial equations (PDEs).

报告人简介：Xinliang Liu is currently a postdoctoral researcher in the Computer, Electrical, and Mathematical Science and Engineering Division (CEMSE) at King Abdullah University of Science and Technology (KAUST), supervised by Prof. Jin-chao Xu. His research focuses on operator learning, multi-scale algorithms and graph nerual networks. He earned his Ph.D. in Computational Mathematics, su-pervised by Prof. Lei Zhang, from Shanghai Jiao Tong University, Shanghai, China, in 2021. Subsequently, he joined KAUST in March 2022.