제목: Neural Operators in Fluid Dynamics: From Turbulence Parameterization to Data Assimilation​​​​

​​

​​​​​​​​​​​​​​​​​​연사: Prof. Heng Xiao (University of Stuttgart)​​

​

장소: 과학관 B102​호​​

Abstract:

​​​Neural operators are emerging tools that learn mappings between physical fields— Abstract such as wind velocity and temperature—directly from simulation or observational data. In this talk, I present how they can be used to model complex, nonlocal processes in turbulent flows and to enable data assimilation by learning mappings between prior and posterior probability distributions. We demonstrate applications in turbulence parameterization using graph neural networks, and in data assimilation using neural operator–based ensemble filters. This approach offers a unified, mesh-independent and sample-efficient framework for learning physical relationships from data.​