{"id":"arxiv/2601.20407","url":"https://arxiv.org/abs/2601.20407","tags":["math.NA"],"score":0,"title":"Refined rates of convergence for target-data dependent greedy generalized interpolation with Sobolev kernels","domain":"math.NA","source":"arxiv","authors":["Bernard Haasdonk","Gabriele Santin","Tizian Wenzel","Daniel Winkle"],"reviews":0,"subject":"math.NA","summary":"Greedy methods have recently been successfully applied to generalized kernel interpolation, or the recovery of a function from data stemming from the evaluation of linear functionals, including the approximation of solutions of linear PDEs by symmetric collocation. When applied to kernels generating Sobolev spaces as their native Hilbert spaces, some of these greedy methods can provide the same error guarantee of generalized interpolation on quasi-uniform points. More importantly, certain target-data-adaptive methods even give a dimension- and smoothness-independent improvement in the speed…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-01-28","reactions":{}}