{"id":"arxiv/2601.20207","url":"https://arxiv.org/abs/2601.20207","tags":["math.NA"],"score":0,"title":"Local Regularity Estimation through Sobolev-Scale Norm Profile","domain":"math.NA","source":"arxiv","authors":["Xiaobin Li","Leevan Ling","Yizhong Sun"],"reviews":0,"subject":"math.NA","summary":"We develop a kernel-based approach for estimating the spatially varying Sobolev regularity~$s$ of an unknown $d$-variate function~$f$ from scattered sampling data, which quantifies the degree of local differentiability supported by the data. Relying only on neighborhood data near the point of interest $z\\in Ω_z$, our method constructs a sequence of Sobolev-space reproducing kernel interpolants whose kernel smoothness order is specified by an index~$m > d/2$. The native-space norms of these interpolants are evaluated over a bounded range of~$m$, producing a \\emph{Sobolev-scale norm profile}.…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-01-28","reactions":{}}