Pith. sign in

REVIEW 2 major objections 6 minor 26 references

Piecewise linear interpolation via kernels

T0 review · 2 major / 6 minor · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper establishes that piecewise linear interpolation is a reproducing-kernel Hilbert space projection: for any positive-definite boundary-weight matrix A and β>0, the Sobolev space W₂¹(0,1) with inner product (2) has a 2-piecewise lin

desk verdict The kernel–spline identification is real and worth knowing; the boundary-condition reduction in Proposition 9 is flawed, so the θ>3/2 superconvergence theorem is not established as written. read the letter →

arxiv 2603.01555 v2 pith:D2QLRMXJ submitted 2026-03-02 math.NA cs.NA

Toni Karvonen , Gabriele Santin , Tizian Wenzel This is my paper · ORCID
classification math.NAcs.NA MSC 41A1565D0565D0746E22
keywords piecewiselinearinterpolationreproducingkernelHilbertspaceGreensuperconvergencefractionalSobolevspacestrapezoidalrule
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper shows that piecewise linear interpolation is not just a classical spline tool but an instance of kernel interpolation. When W₂¹(0,1) is equipped with an inner product formed from a positive-definite matrix of boundary terms plus a derivative term, the reproducing kernel is exactly a 2-piecewise linear function, so kernel interpolation at the nodes produces the linear spline interpolant. The same kernels are Green functions of a two-point boundary value problem, which lets the authors apply kernel-based superconvergence theory to derive L² error bounds of order h^θ for functions in W₂^θ(0,1), θ∈[1,2]. These rates coincide with classical linear-spline rates, so the paper unifies two previously separate approximation theories. If correct, it gives a single framework that explains both the approximation and quadrature (trapezoidal rule) behavior of linear splines.

What carries the argument

The machinery is the bilinear form (2), parameterized by a 2×2 boundary matrix A and a scale β, which turns W₂¹(0,1) into an RKHS with explicit kernel (3). The kernel's piecewise-linearity forces kernel interpolants to be linear splines (Theorem 2). The Green-function representation (11) makes the kernel the integral kernel of a second-order differential operator, so the power spaces H_θ — defined by real interpolation between the RKHS and T L₂ — can be characterized via elliptic regularity and Triebel's interpolation theorem for normal boundary systems. The final ingredient is the general superconvergence estimate (Theorem 6), which converts the order-h sampling bound for H into order-h^θ b

What would settle it

Pick α₀=α₁=1, α₂=1/2, β=1, nodes xᵢ=i/n, and f(x)=x^{5/2} (in W₂^{1.8} but not satisfying (11)). Compute the L² error of the piecewise linear interpolant for n=10,20,40 and compare the empirical rate to θ=1.8 from Corollary 10. If the rate is closer to 1.5 or degrades, the power-space identification in Proposition 9 is wrong for α₂≠0. Also verify directly whether the kernel in (3) equals the Green function of (11) by checking the boundary residuals βu′(0)−α₀u(0)−α₂u(1).

Watch

Extended reading notes

Core claim

The central claim is that the reproducing kernel of the Sobolev space W₂¹(0,1) under the inner product ⟨f,g⟩ = α₀f(0)g(0) + α₁f(1)g(1) + α₂(f(0)g(1)+f(1)g(0)) + β⟨f′,g′⟩_{L₂} is 2-piecewise linear whenever β>0 and the boundary matrix A is positive-definite. Consequently the kernel interpolant equals the piecewise linear interpolant for every function in the space. Moreover this kernel is the Green function of the second-order boundary value problem −βu″=f with boundary conditions βu′(0)=α₀u(0)+α₂u(1) and βu′(1)=−α₁u(1)−α₂u(0). Using interpolation-space and superconvergence arguments, the authors identify the associated power spaces H_θ and obtain the bound ‖u−Pₙu‖_{L₂} ≤ (1/(√2 β))^θ h^θ ‖u‖

Load-bearing premise

The paper's superconvergence rates for θ>3/2 depend on the claim that the boundary conditions (11) form a 'normal system' that can be rewritten as two local differential equations at the endpoints; for α₂≠0 this rewriting forces additional constraints such as α₂u(1)=0 that are not part of (11), so if that reduction is invalid the identification of H_θ and the θ>3/2 bound lack support.

Editorial extensions

If this is right

  • Kernel interpolation with any of the kernels (3) is exactly piecewise linear interpolation, so the two perspectives share all optimality and worst-case properties.
  • The Green-kernel identification means the linear spline interpolant solves a penalized boundary-value problem; the parameters α₀, α₁, α₂, β control the boundary behavior of the interpolant.
  • The superconvergence theorem gives L² error O(h^θ) for u in W₂^θ(0,1), θ∈[1,2] — the same rates as classical spline theory — with explicit constants depending only on β and θ.
  • Kernel quadrature for these kernels coincides with the trapezoidal rule, so the paper also yields a superconvergence-style proof of trapezoidal-rule error estimates.
  • For θ>3/2 the error bound requires the interpolated function to satisfy the boundary conditions (11), reflecting the fact that the kernel's native space reaches only part of W₂².

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • This framework likely extends to higher-order splines: choosing a Sobolev inner product with boundary terms whose kernel is a 2m-piecewise polynomial of degree m should make kernel interpolation coincide with spline interpolation of order m, with superconvergence supplying the rates.
  • The explicit kernel formula gives a ready-made covariance model for Bayesian or probabilistic numerical methods on [0,1] whose posterior mean is the linear spline, potentially simplifying uncertainty quantification for spline smoothing.
  • The superconvergence machinery could be tested at the critical exponents θ=1/2 and θ=3/2, where Proposition 9 is silent, by direct numerical rate computations on finite meshes.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 6 minor

Summary. The paper studies piecewise linear interpolation on [0,1] from the RKHS point of view. For the inner product (2) on W_2^1(0,1) (with β>0 and the matrix A positive definite), it derives the explicit reproducing kernel (3), shows that this kernel is 2-piecewise linear so that kernel interpolation coincides with linear spline interpolation (Theorem 2), and proves that the kernel is the Green kernel of the second-order nonlocal boundary value problem (11) (Corollary 5). It then applies a general superconvergence theorem from the authors' preprint [13] (Theorem 6) and a sampling inequality (Proposition 7) to obtain L2 error bounds of order h^θ for θ∈[1,2]. The higher-regularity range θ>3/2 relies on an identification of the interpolation spaces H_θ in Proposition 9. Corollary 10 states the resulting explicit constants. The last section reviews classical spline and trapezoidal-rule error bounds.

Significance. The kernel–spline identification is attractive and the Green-kernel calculation is a concrete, useful contribution. The paper contains no fitted parameters and the main objects are explicit. If the interpolation-space identification can be made rigorous, the superconvergence route to classical linear-spline rates would provide a nice conceptual bridge. However, the manuscript as written contains a false constant in the main convergence bound and an invalid reduction of the nonlocal boundary conditions to local normal boundary operators. The claimed higher-regularity bounds are therefore not established as written.

major comments (2)
  1. [§4, Corollary 10 and the note before it] From (2) one has |f|_{W_2^1}^2 ≤ β^{-1} ||f||_H^2, so the embedding constant needed in Corollary 8 is c = β^{-1/2}, not 1/β. The resulting bound in Corollary 10 should be (1/√(2β))^θ h^θ ||u||_{H_θ}, not (1/(√2 β))^θ h^θ ||u||_{H_θ}. For β>1 the stated constant is too small; e.g. with β=4, α0=α1=1, α2=0 and f(x)=x, |f|_{W_2^1}=1 and ||f||_H^2=5, so the ratio is 1/√5, whereas 1/β = 1/4 would give a false upper bound. Since Corollary 10 is the paper's main quantitative convergence claim, this is a load-bearing error.
  2. [§4, Proposition 9 (proof of the boundary-operator reduction)] The reduction of (11) to (B_j u)|_{∂Ω}=0 is not valid. With the coefficients stated in the proof, B_1u(1) = -α2 u(1) and B_2u(0) = α1 u(0). Requiring all four values (B_j u)(0) and (B_j u)(1) to vanish therefore forces α1 u(0)=0 and α2 u(1)=0, which are not part of (11). The boundary conditions in (11) are genuinely nonlocal: the left condition contains u(1) and the right condition contains u(0), so no pair of local differential operators at the endpoints can represent them. A concrete counterexample is α0=α1=1, α2=1/2, β=1, u(x)=-3/2 x^2 + 3/2 x + 1, which satisfies (11) but has B_1u(1)=-1/2 and B_2u(0)=1. Hence the claimed equality H_2 = {u∈W_2^2 : (B_j u)|_{∂Ω}=0} is false, the Triebel normal-system interpolation theorem is inapplicable, and the characterization of H_θ for θ>3/2 — together with Corollary 10 in that range — is unsupported.
minor comments (6)
  1. [§4, Proposition 9] The statement says θ∈[1,2]\{1/2}, but the critical case omitted by the interpolation argument is θ=3/2, not 1/2. The proof also refers to 'Theorem 5' where Corollary 5 is meant.
  2. [§3, Corollary 5] The hypotheses say 'Under the assumptions of Theorem 2', but the kernel formula and boundary value problem require the assumptions of Theorem 3.
  3. [Throughout] Several internal references are mismatched: the paragraph after Proposition 7 refers to 'Theorem 7'; the note after Corollary 8 refers to 'Theorem 8'; and the paragraph after Corollary 10 refers to 'Theorem 10'.
  4. [§5.1, Theorem 11] Two parts are labelled (d). The second (d) should be (e), and the following Besov item should be relabelled accordingly.
  5. [Abstract and Introduction] The notation W_2^s(0,1) is used in the abstract before the Sobolev spaces are defined in Section 2. There are also minor typos ('streches', 'Furhermore').
  6. [§4, Theorem 6] The superconvergence theorem is imported without proof from the authors' own arXiv preprint [13]. Since it is the engine of the convergence argument, please either include a proof of the needed statement or update the reference to a published version.

Circularity Check

1 steps flagged · score 2.0 of 10

No significant circularity: the kernel–spline identification and Green-kernel steps are derived from first principles; only the superconvergence transfer theorem is imported from the authors' own preprint [13].

  1. self citation load bearing [Section 4, Theorem 6 and proof of Proposition 9]
    "The following general superconvergence theorem follows from [13, Cor. 15]. ... Finally, Hθ = (L2(0,1), T L2(0,1)) θ/2,2 for all θ∈(1,2) with equivalent norms [13]."

    The central rate-transfer mechanism is not proved in the present paper but cited to [13], a preprint by the same three authors. Corollaries 8 and 10 inherit this theorem. However, [13] is a general interpolation-space result, not a statement tuned to piecewise linear kernels or to the classical spline rates, so this is a load-bearing self-citation rather than a circular reduction of the target result. It therefore raises the score only mildly.

full rationale

The main derivation chain is self-contained: Theorem 3 constructs the inner product and solves for the reproducing kernel explicitly from the reproducing property; Theorem 2 and Corollary 5 identify the kernel interpolant with the linear spline and with the Green kernel of the stated PDE by direct calculation; Proposition 7 proves the sampling inequality; and Proposition 9 invokes external interpolation theory (Triebel) in addition to the authors' preprint. No fitted parameter is renamed as a prediction, and the claimed convergence rates are not put into the construction. The only load-bearing self-citation is [13] supplying the general superconvergence theorem used in Theorem 6/Corollary 8; because it is a general transfer principle and not an ansatz that encodes the target rates, the circularity burden is low. Note that the proof of Proposition 9's local boundary-operator reformulation of (11) appears mathematically questionable (it forces u(0)=u(1)=0 when α1α2≠0, which is not part of (11)); that is a correctness/validity concern, not a circularity, and is not counted in the score.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

This paper does not fit any data; the constants α0,α1,α2,β parametrize the kernel family and are free symbolic inputs, not fitted parameters. The derivation rests on standard RKHS theory, Green-function facts, the authors' prior superconvergence theorem [13], and Triebel's interpolation characterization.

assumptions (5)
  • standard math Moore–Aronszajn theorem and basic RKHS interpolation theory (kernel interpolant is the orthogonal projection and worst-case optimal)
    Invoked in Section 1 to define kernel interpolation and its optimality properties.
  • standard math W2^1(0,1) embeds continuously into C[0,1]; point evaluations are bounded; the norm induced by (2) is equivalent to the standard Sobolev norm
    Used in the proof of Theorem 3 to show (2) is an inner product whose completion is an RKHS.
  • domain assumption Superconvergence theorem [13, Cor. 15]: a global O(ε) error bound on the native space implies O(ε^θ) on the interpolation spaces H_θ
    Not proved in this paper; imported from the authors' related preprint (arXiv:2505.11435). Load-bearing for Corollaries 8 and 10.
  • domain assumption Triebel's interpolation characterization of domains of elliptic operators with normal boundary systems ([23, Thm 1, Sec. 4.3.3])
    Used in Proposition 9 to identify H_θ with W2^θ(0,1), with boundary conditions only when θ>3/2.
  • standard math Green-kernel characterization via piecewise affinity, jump condition, and boundary conditions ([8, Sec. 6.2–6.4])
    Used in the proof of Corollary 5 to identify the reproducing kernel with the Green kernel of (11).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Piecewise linear interpolation via kernels." pith.science (2026). https://pith.science/paper/D2QLRMXJ

@misc{pith2026260301555,
  author       = {Pith},
  title        = {Pith review of: Piecewise linear interpolation via kernels},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D2QLRMXJ}},
  note         = {Machine review of arXiv:2603.01555}
}
abstract

We consider piecewise linear interpolation from the perspective of kernel interpolation and quadrature. If the Sobolev space $W_2^1(0, 1)$ is equipped with a suitable inner product, its reproducing kernel is piecewise linear and gives rise to piecewise linear interpolation. We show that such kernels are Green kernels for certain second-order partial differential equations and use kernel-based superconvergence theory to obtain rates of convergence for approximation of functions lying in $W_2^s(0, 1)$ for $s \in [1, 2]$. The rates coincide with classical rates for linear splines.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 1 linked inside Pith

  1. [13]

    Karvonen, G

    T. Karvonen, G. Santin, and T. Wenzel. General superconvergence for kernel-based approximation.arXiv:2505.11435, 2025

    arXiv 2025

  2. [1]

    Arcang´ eli, M

    R. Arcang´ eli, M. C. L´ opez de Silanes, and J. J. Torrens. An extension of a bound for functions in sobolev spaces, with applications to (m, s)- spline interpolation and smoothing.Numerische Mathematik, 107(2):181– 211, Aug 2007

    2007

  3. [2]

    Aronszajn

    N. Aronszajn. Theory of reproducing kernels.Transactions of the American Mathematical Society, 68(3):337–404, 1950

    1950

  4. [3]

    Berlinet and C

    A. Berlinet and C. Thomas-Agnan.Reproducing Kernel Hilbert Spaces in Probability and Statistics. Springer, 2004

    2004

  5. [4]

    Cruz-Uribe and C

    D. Cruz-Uribe and C. J. Neugebauer. Sharp error bounds for the trape- zoidal rule and Simpson’s rule.Journal of Inequalities in Pure and Applied Mathematics, 3(4), 2002. Article 49

    2002

  6. [5]

    Diaconis

    P. Diaconis. Bayesian numerical analysis. InStatistical decision theory and related topics IV, volume 1, pages 163–175. Springer-Verlag New York, 1988

    1988

  7. [6]

    S. S. Dragomir and S. Mabizela. Some error estimates in the trapezoidal quadrature rule.RGMIA Research Report Collection, 5(2), 1999

    1999

  8. [7]

    Duc-Jacquet.Approximation des fonctionnelles lin´ eaires sur les es- paces hilbertiens autoreproduisants

    M. Duc-Jacquet.Approximation des fonctionnelles lin´ eaires sur les es- paces hilbertiens autoreproduisants. PhD thesis, Universit´ e Joseph-Fourier- Grenoble I, 1973

    1973

Show all 26 references
  1. [8]

    G. E. Fasshauer and M. J. McCourt.Kernel-based Approximation Methods using MATLAB. World Scientific Publishing, 2015

    2015

  2. [9]

    G. E. Fasshauer and Q. Ye. Reproducing kernels of generalized Sobolev spaces via a Green function approach with distributional operators.Nu- merische Mathematik, 119:585–611, 2011

    2011

  3. [10]

    G. E. Fasshauer and Q. Ye. Reproducing kernels of Sobolev spaces via a Green kernel approach with differential operators and boundary operators. Advances in Computational Mathematics, 38:891–921, 2013

    2013

  4. [11]

    G. W. Hedstrom and R. S. Varga. Application of Besov spaces to spline approximation.Journal of Approximation Theory, 4(3):295–327, 1971

    1971

  5. [12]

    Iske.Approximation Theory and Algorithms for Data Analysis

    A. Iske.Approximation Theory and Algorithms for Data Analysis. Springer, Cham, 2018. 11

    2018

  6. [14]

    Novak and H

    E. Novak and H. Wo´ zniakowski.Tractability of Multivariate Problems. Volume II: Standard Information for Functionals. European Mathematical Society, 2010

    2010

  7. [15]

    V. I. Paulsen and M. Raghupathi.An Introduction to the Theory of Re- producing Kernel Hilbert Spaces. Cambridge University Press, 2016

    2016

  8. [16]

    Ritter.Average-Case Analysis of Numerical Problems

    K. Ritter.Average-Case Analysis of Numerical Problems. Springer, 2000

    2000

  9. [17]

    Santin, T

    G. Santin, T. Karvonen, and B. Haasdonk. Sampling based approximation of linear functionals in reproducing kernel Hilbert spaces.BIT Numerical Mathematics, 62:279–310, 2021

    2021

  10. [18]

    Schaback

    R. Schaback. Superconvergence of kernel-based interpolation.Journal of Approximation Theory, 235:1–19, 2018

    2018

  11. [19]

    I. H. Sloan and V. Kaarnioja. Doubling the rate: improved error bounds for orthogonal projection with application to interpolation.BIT Numerical Mathematics, 65(1):10, 2025

    2025

  12. [20]

    A. V. Suldin. Wiener measure and its applications to approximation meth- ods. I.Izvestiya vysshikh uchebnykh zavedeniy. Matematika, (6):145–158, 1959

    1959

  13. [21]

    A. V. Suldin. Wiener measure and its applications to approximation meth- ods. II.Izvestiya vysshikh uchebnykh zavedeniy. Matematika, (5):165–179, 1960

    1960

  14. [22]

    B. L. Swartz and R. S. Varga. Error bounds for spline andL-spline inter- polation.Journal of Approximation Theory, 6(1):6–49, 1972

    1972

  15. [23]

    Triebel.Interpolation Theory, Function Spaces, Differential Operators

    H. Triebel.Interpolation Theory, Function Spaces, Differential Operators. North-Holland, 1978

    1978

  16. [24]

    Wahba.Spline Models for Observational Data

    G. Wahba.Spline Models for Observational Data. Society for Industrial and Applied Mathematics, 1990

    1990

  17. [25]

    Wendland.Scattered Data Approximation

    H. Wendland.Scattered Data Approximation. Cambridge University Press, 2005

    2005

  18. [26]

    Wendland and C

    H. Wendland and C. Rieger. Approximate interpolation with applications to selecting smoothing parameters.Numerische Mathematik, 101(4):729– 748, 2005. 12

    2005

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.