{"id":"arxiv/2602.00870","url":"https://arxiv.org/abs/2602.00870","tags":["math.NA"],"score":0,"title":"Finite Element Eigenfunction Network (FEENet): A Hybrid Framework for Solving PDEs on Complex Geometries","domain":"math.NA","source":"arxiv","authors":["Shiyuan Li","Hossein Salahshoor"],"reviews":0,"subject":"math.NA","summary":"Neural operators aim to learn mappings between infinite-dimensional function spaces, but their performance often degrades on complex or irregular geometries due to the lack of geometry-aware representations. We propose the Finite Element Eigenfunction Network (FEENet), a hybrid spectral learning framework grounded in the eigenfunction theory of differential operators. For a given domain, FEENet leverages the Finite Element Method (FEM)toperformaone-timecomputationofaneigenfunctionbasisintrinsictothegeometry. PDE solutions are subsequently represented in this geometry-adapted basis, and…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-01-31","reactions":{}}