Solving Partial Differential Equations Based on Unstructured Neural Operators
ID:15
View Protection:ATTENDEE
Updated Time:2025-09-30 10:34:56
Hits:410
Oral Presentation
Abstract
We present a neural operator that directly handles unstructured CFD data via projection-based attention. It captures complex spatial dependencies and generalizes across geometries. Numerical experiments on structured and unstructured PDE datasets show superior accuracy and efficiency compared with existing approaches.
Keywords
operator learning,unstructured grid,flow and heat transfer
Submission Author
Haobo Guo
Xi'An Jiaotong University
Wen-Quan Tao
Xi'an Jiaotong University
Submit Comment