报告题目：Generative downscaling of PDE solvers with physics-guided diffusion models

报告时间：2024-06-12 10:50-11:50

报  告 人：徐悟哲  博士（UMass Amherst）

报告地点：理学院东北楼四楼报告厅(404)

Abstract: In this talk, I will introduce our novel Physics-Guided Diffusion Model (PGDM) for solving complex PDEs via downscaling, transforming low-resolution data into high-fidelity solutions. Our numerical experiments demonstrate that PGDM not only achieves computation speeds over ten times faster than traditional methods but also exceeds the accuracy of other data-driven approaches like FNO, particularly in limiting data scenarios.