{"id":"arxiv/2607.21287","url":"https://arxiv.org/abs/2607.21287","tags":["math.NA"],"score":0,"title":"Adaptive Non-Linear Partition of Unity Methods for Scattered Data Interpolation with Discontinuities","domain":"math.NA","source":"arxiv","authors":["Adeeba Haider","Roberto Cavoretto","Juan Ruiz-Álvarez","Dionisio F. Yáñez"],"reviews":0,"subject":"math.NA","summary":"Scattered data approximation with discontinuities is challenging due to the Gibbs phenomenon, which significantly reduces accuracy near interfaces. The recently introduced Non-Linear Partition of Unity Method (NL-PUM) addresses this by combining Radial Basis Function (RBF) interpolation with a non-linear Weighted Essentially Non-Oscillatory (WENO) strategy. While effective, NL-PUM's performance relies heavily on two fixed hyperparameters: the RBF shape parameter and the patch radius. This work extends NL-PUM by adapting both hyperparameters locally using Leave-One-Out Cross-Validation…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-07-23","reactions":{}}