pith. sign in

arxiv: 2606.28564 · v1 · pith:AWYRVDCOnew · submitted 2026-06-26 · 🧮 math.CA · cs.NA· math.NA

Kernel approximation beyond the native space -- with applications to approximation on manifolds

Thomas Hangelbroek , Christian Rieger , Grady B. Wright This is my paper

Pith reviewed 2026-06-30 00:59 UTC · model grok-4.3

classification 🧮 math.CA cs.NAmath.NA
keywords kernel approximationembedded manifoldsSobolev spacesintegral operatorsBernstein inequalitiesinterpolationnative spacepositive definite kernels
0
0 comments X

The pith

Kernel approximation on manifolds extends to Sobolev spaces smoother than the native space.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper generalizes an approximation scheme from DeVore and Ron to embedded manifolds, targeting functions in the range of the kernel's integral operator instead of the native space. This sidesteps the failure of standard RKHS orthogonality arguments when error is measured in higher-regularity spaces. Sufficient conditions are given to identify that range precisely as a Sobolev space. New Bernstein inequalities for kernels on manifolds then yield interpolation error bounds in Sobolev spaces compactly contained in the native space.

Core claim

By using local polynomial reproductions on submanifolds of R^N, the range of the integral operator associated to positive or conditionally positive definite kernels is identified with a Sobolev space; this identification permits the DeVore-Ron scheme to produce approximation rates and, together with new kernel-based Bernstein inequalities, to control interpolation errors in Sobolev spaces that sit strictly inside the native space.

What carries the argument

The range of the kernel integral operator, identified as a Sobolev space on the manifold via local polynomial reproductions.

If this is right

  • Error estimates hold for interpolation in Sobolev spaces compactly contained in the native space.
  • New Bernstein inequalities are available for restrictions of kernels to embedded manifolds.
  • The range of the integral operator coincides with a Sobolev space under stated conditions on kernel and manifold.
  • The scheme applies to both positive definite and conditionally positive definite kernels.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Kernel methods could therefore handle targets whose smoothness exceeds that of the chosen kernel without requiring a change of kernel.
  • The spectral properties of the integral operator on the manifold become directly usable for rate calculations.
  • The same local-reproduction technique may transfer to approximation on other geometries where polynomial reproduction holds locally.

Load-bearing premise

Local polynomial reproductions exist for the given embedded submanifolds, allowing the DeVore-Ron scheme to be carried over.

What would settle it

A concrete numerical example on a manifold where the observed approximation rate for a target in a higher Sobolev space fails to match the rate predicted by membership in the integral-operator range.

Figures

Figures reproduced from arXiv: 2606.28564 by Christian Rieger, Grady B. Wright, Thomas Hangelbroek.

Figure 1
Figure 1. Figure 1: Target function (5.6) used in the numerical experiments. [PITH_FULL_IMAGE:figures/full_fig_p023_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Plots of (1 + ℓ(ℓ + 1))τ Pℓ µ=−ℓ |IdΞf(ℓ, µ) − fb(ℓ, µ)| 2 for various N = |Ξ|. (a)–(d) Show the results using different τ . The solid black line corresponds to the line ℓ −2(4−τ)+1 and give a measure of the slopes of the tails of the coefficients that are summed to approximate the Hτ (S 2 ) norm In computing these results, we approximated the Sobolev norm (5.7) by truncating the sum to degree L = 1024. Th… view at source ↗
read the original abstract

This article treats kernel approximation and interpolation on embedded manifolds of $\mathbb{R}^N$using restrictions of positive and conditionally positive definite kernels. The main challenge is to develop an approximation theory that treats error measured in highly regular smoothness spaces relative to the kernel. This means that the order of smoothness is higher than that of the kernel's associated native space (in the positive definite case, the reproducing kernel Hilbert space generated by the kernel). This prevents the use of standard techniques for controlling error in this setting, especially RKHS space arguments like orthogonality of the interpolation projector, or bounds using the {\em power function}. We generalize an approximation scheme introduced by DeVore and Ron which treats target functions that are in the range of the kernel's integral operator. In the case of embedded manifolds, this generalization is now feasible due to recently developed local polynomial reproductions for certain submanifolds of $\mathbb{R}^N$. Furthermore, we give sufficient conditions on kernel and manifold which allow the range of the integral operator to be precisely identified: in particular, guaranteeing that the range is a Sobolev space. Finally, we provide new kernel-based Bernstein inequalities for embedded manifolds which lead to estimates for interpolation in Sobolev spaces compactly contained in the native space.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The paper generalizes the DeVore-Ron approximation scheme to kernel approximation and interpolation on embedded manifolds of R^N, using restrictions of positive and conditionally positive definite kernels. It addresses error estimates in smoothness spaces of higher order than the native space by considering target functions in the range of the kernel's integral operator, made feasible by recent local polynomial reproduction results on submanifolds. The work supplies kernel and manifold conditions under which this range is identified with a Sobolev space, and derives new kernel-based Bernstein inequalities that yield interpolation error estimates in Sobolev spaces compactly embedded in the native space.

Significance. If the central claims hold, the results meaningfully extend kernel approximation theory into the regime of target functions smoother than the native space, a setting where standard RKHS orthogonality and power-function arguments fail. By leveraging external local polynomial reproduction theorems and providing explicit range-identification conditions that recover Sobolev spaces, the manuscript supplies a coherent framework for high-order approximation on manifolds. The new Bernstein inequalities are a concrete technical contribution that directly enables the interpolation estimates. These tools are likely to be useful in geometric approximation and manifold-based numerical analysis.

minor comments (2)
  1. [Abstract] The abstract states that the generalization 'is now feasible due to recently developed local polynomial reproductions' but does not name the specific references; the introduction should cite the precise works on local polynomial reproduction for submanifolds of R^N.
  2. [Introduction] The notation for the integral operator, its range, and the associated Sobolev identification is introduced gradually; a dedicated preliminary subsection collecting the operator definitions, the DeVore-Ron scheme, and the target Sobolev spaces would improve readability for readers outside the immediate subfield.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the careful reading and the positive assessment of the manuscript's contributions. The recommendation for minor revision is appreciated, and we will incorporate improvements to clarity and presentation in the revised version. No specific major comments were raised in the report.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper's core program generalizes the external DeVore-Ron scheme for functions in the range of the integral operator, invokes recently developed external local polynomial reproduction results to enable the manifold case, states sufficient kernel/manifold conditions to identify the range as a Sobolev space, and derives new Bernstein inequalities. No load-bearing step reduces by construction to a self-defined quantity, a fitted input renamed as prediction, or a self-citation chain; all cited foundations are external and the derivation remains self-contained against those benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only; no explicit free parameters, invented entities, or detailed axioms listed. The key enabling assumption is the existence of local polynomial reproductions.

axioms (1)
  • domain assumption Recently developed local polynomial reproductions exist for certain submanifolds of R^N
    Cited as the reason the generalization is now feasible.

pith-pipeline@v0.9.1-grok · 5759 in / 1142 out tokens · 32370 ms · 2026-06-30T00:59:19.990942+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

34 extracted references · 1 canonical work pages

  1. [1]

    Stegun.Handbook of mathematical functions with formu- las, graphs, and mathematical tables

    Milton Abramowitz and Irene A. Stegun.Handbook of mathematical functions with formu- las, graphs, and mathematical tables. National Bureau of Standards Applied Mathematics Series, No. 55. U. S. Government Printing Office, Washington, D.C., 1964. For sale by the Superintendent of Documents

    1964

  2. [2]

    Charles R. Baker. Complete simultaneous reduction of covariance operators.SIAM J. Appl. Math., 17:972–983, 1969. 24

    1969

  3. [3]

    Brad J. C. Baxter and Simon Hubbert. Radial basis functions for the sphere. InRe- cent progress in multivariate approximation (Witten-Bommerholz, 2000), volume 137 of Internat. Ser. Numer. Math., pages 33–47. Birkhäuser, Basel, 2001

    2000

  4. [4]

    An introduction, volume No

    Jöran Bergh and Jörgen Löfström.Interpolation spaces. An introduction, volume No. 223 ofGrundlehren der Mathematischen Wissenschaften. Springer-Verlag, Berlin-New York, 1976

    1976

  5. [5]

    J. H. Bramble and S. R. Hilbert. Estimation of linear functionals on Sobolev spaces with application to Fourier transforms and spline interpolation.SIAM J. Numer. Anal., 7:112– 124, 1970

    1970

  6. [6]

    On the mathematical foundations of learning.Bull

    Felipe Cucker and Steve Smale. On the mathematical foundations of learning.Bull. Amer. Math. Soc. (N.S.), 39(1):1–49, 2002

    2002

  7. [7]

    Approximation using scattered shifts of a multivariate function.Trans

    Ronald DeVore and Amos Ron. Approximation using scattered shifts of a multivariate function.Trans. Amer. Math. Soc., 362(12):6205–6229, 2010

    2010

  8. [8]

    The Green formula and layer potentials.Integral Equations Operator Theory, 41(2):127–178, 2001

    Roland Duduchava. The Green formula and layer potentials.Integral Equations Operator Theory, 41(2):127–178, 2001

    2001

  9. [9]

    Edward Fuselier and Grady B. Wright. Scattered data interpolation on embedded subman- ifolds with restricted positive definite kernels: Sobolev error estimates.SIAM J. Numer. Anal., 50(3):1753–1776, 2012

    2012

  10. [10]

    Springer, New York, third edition, 2014

    Loukas Grafakos.Modern Fourier analysis, volume 250 ofGraduate Texts in Mathematics. Springer, New York, third edition, 2014

    2014

  11. [11]

    Springer, New York, 2009

    Gerd Grubb.Distributions and operators, volume 252 ofGraduate Texts in Mathematics. Springer, New York, 2009

    2009

  12. [12]

    Springer, Heidelberg, 2015

    WolfgangHackbusch.Hierarchical matrices: algorithms and analysis, volume49ofSpringer Series in Computational Mathematics. Springer, Heidelberg, 2015

    2015

  13. [13]

    Hangelbroek and C

    T. Hangelbroek and C. Rieger. Extending error bounds for radial basis function interpola- tion to measuring the error in higher order Sobolev norms.Math. Comp., 94(351):381–407, 2025

    2025

  14. [14]

    Polyharmonic and related kernels on manifolds: interpolation and approximation.Foundations of Computational Mathematics, 12:625–670, 2012

    Thomas Hangelbroek, Francis J Narcowich, and Joseph D Ward. Polyharmonic and related kernels on manifolds: interpolation and approximation.Foundations of Computational Mathematics, 12:625–670, 2012

    2012

  15. [15]

    Generalized local polynomial reproductions.arXiv preprint arXiv:2410.12973, 2024

    Thomas Hangelbroek, Christian Rieger, and Grady B Wright. Generalized local polynomial reproductions.arXiv preprint arXiv:2410.12973, 2024

    work page arXiv 2024

  16. [16]

    Hangelbroek

    Thomas C. Hangelbroek. On a polyharmonic Dirichlet problem and boundary effects in surface spline approximation.SIAM J. Math. Anal., 50(4):4616–4654, 2018

    2018

  17. [17]

    Harbrecht, M

    H. Harbrecht, M. Multerer, O. Schenk, and Ch. Schwab. Multiresolution kernel matrix algebra.Numer. Math., 156(3):1085–1114, 2024

    2024

  18. [18]

    Hubbert and B

    S. Hubbert and B. Baxter. Radial basis functions for the sphere. In W. Haussmann, K. Jetter, and M. Reimer, editors,Recent Progress in Multivariate Approximation, Proc. of the 4th Intern. Conf., Witten-Bommerholz, Germany, volume 137 ofInternational Series of Numerical Mathematics, Basel, 2001. Birkhäuser

    2001

  19. [19]

    Springer, 2015

    Simon Hubbert, Quôc Thông Lê Gia, Tanya M Morton, et al.Spherical radial basis func- tions, theory and applications. Springer, 2015

    2015

  20. [20]

    Johnen and K

    H. Johnen and K. Scherer. On the equivalence of theK-functional and moduli of continuity and some applications. InConstructive theory of functions of several variables (Proc. Conf., Math. Res. Inst., Oberwolfach, 1976), volume Vol. 571 ofLecture Notes in Math., pages 119–140. Springer, Berlin-New York, 1977

    1976

  21. [21]

    Jonsson and H

    A. Jonsson and H. Wallin. A Whitney extension theorem inLp and Besov spaces.Ann. Inst. Fourier (Grenoble), 28(1):vi, 139–192, 1978. 25

    1978

  22. [22]

    Tempered radon measures.Revista Matemática Complutense, 21(2):553–564, jul

    Maryia Kabanava. Tempered radon measures.Revista Matemática Complutense, 21(2):553–564, jul. 2008

    2008

  23. [23]

    Hilbert-Schmidt regularity of sym- metric integral operators on bounded domains with applications to SPDE approximations

    Mihály Kovács, Annika Lang, and Andreas Petersson. Hilbert-Schmidt regularity of sym- metric integral operators on bounded domains with applications to SPDE approximations. Stoch. Anal. Appl., 41(3):564–590, 2023

    2023

  24. [24]

    Function recovery on manifolds using scattered data

    David Krieg and Mathias Sonnleitner. Function recovery on manifolds using scattered data. Journal of Approximation Theory, 305:106098, 2025

    2025

  25. [25]

    Distributed learning with regularized least squares.J

    Shao-Bo Lin, Xin Guo, and Ding-Xuan Zhou. Distributed learning with regularized least squares.J. Mach. Learn. Res., 18:Paper No. 92, 31, 2017

    2017

  26. [26]

    Odell and J

    C. Odell and J. Levesley. Evaluation of some integrals arising from approximation on the sphere using radial basis functions.Numer. Funct. Anal. Optim., 23(3-4):359–365, 2002

    2002

  27. [27]

    Schaback

    R. Schaback. Improved error bounds for scattered data interpolation by radial basis func- tions.Math. Comp., 68(225):201–216, 1999

    1999

  28. [28]

    Efficient spherical harmonic transforms aimed at pseudospectral nu- merical simulations.Geochemistry, Geophysics, Geosystems, 14(3):751–758, 2013

    Nathanaël Schaeffer. Efficient spherical harmonic transforms aimed at pseudospectral nu- merical simulations.Geochemistry, Geophysics, Geosystems, 14(3):751–758, 2013

    2013

  29. [29]

    Sloan and Robert S

    Ian H. Sloan and Robert S. Womersley. Extremal systems of points and numerical integra- tion on the sphere.Advances in Computational Mathematics, 192:107–125, 2004

    2004

  30. [30]

    Mercer’s theorem on general domains: on the interaction between measures, kernels, and RKHSs.Constr

    Ingo Steinwart and Clint Scovel. Mercer’s theorem on general domains: on the interaction between measures, kernels, and RKHSs.Constr. Approx., 35(3):363–417, 2012

    2012

  31. [31]

    A. M. Stuart. Inverse problems: a Bayesian perspective.Acta Numer., 19:451–559, 2010

    2010

  32. [32]

    II, volume 84 ofMonographs in Mathematics

    Hans Triebel.Theory of function spaces. II, volume 84 ofMonographs in Mathematics. Birkhäuser Verlag, Basel, 1992

    1992

  33. [33]

    Local polynomial reproduction and moving least squares approximation

    Holger Wendland. Local polynomial reproduction and moving least squares approximation. IMA J. Numer. Anal., 21(1):285–300, 2001

    2001

  34. [34]

    Cambridge University Press, Cambridge, 2005

    Holger Wendland.Scattered data approximation, volume 17 ofCambridge Monographs on Applied and Computational Mathematics. Cambridge University Press, Cambridge, 2005. 26

    2005