pith. sign in

arxiv: 2604.27160 · v1 · submitted 2026-04-29 · 🧮 math.NA · cs.NA

Embeddings of Reproducing Kernel Hilbert Spaces with General Weights

Michael Gnewuch , Peter Kritzer , Klaus Ritter This is my paper

Pith reviewed 2026-05-07 10:34 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords reproducing kernel Hilbert spacesembeddingsweighted tensor productscompletely monotone weightsdiscrete calculusnumerical integrationfunction recoveryinfinite dimensions
0
0 comments X

The pith

Embeddings between reproducing kernel Hilbert spaces are obtained by transforming weights to offset changes in the univariate kernel.

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

The paper shows that for kernels formed as weighted superpositions of tensor products from one fixed univariate kernel, one space embeds into another when the weights are adjusted to compensate for any change in that univariate kernel. This holds for functions of finitely many or countably infinitely many variables. The proofs rely on a discrete calculus developed on the cone of weight sequences, with completely monotone weights playing a central role in establishing the required norm inequalities. Such embeddings matter for computational tasks because they let researchers transfer error estimates or convergence rates between different kernel choices without starting from scratch each time. The approach therefore supports more flexible design of algorithms for high-dimensional numerical integration and function recovery.

Core claim

For kernels K that are superpositions of weighted finite tensor products of a fixed univariate kernel, an embedding of the associated reproducing kernel Hilbert space H(K) into H(K') holds whenever the weights are transformed in a manner that compensates the change from the original univariate kernel to the new one. For finite dimension d the argument employs a discrete calculus on the cone of all weights; for infinite d the same calculus is extended. Completely monotone weights are singled out as the case in which the embedding constants remain controlled.

What carries the argument

Discrete calculus on the cone of weights, with completely monotone weights serving as the distinguished subclass that guarantees the embedding inequalities after suitable weight transformation.

If this is right

  • Embeddings exist uniformly for both finite and countably infinite numbers of variables.
  • Error bounds derived for one kernel in a weighted setting carry over to a second kernel after the weight transformation.
  • Quadrature rules or recovery algorithms designed for one kernel can be compared directly to those for another kernel via the embedding constants.
  • The calculus supplies an explicit mechanism for constructing the required weight maps between arbitrary pairs of univariate kernels.

Where Pith is reading between the lines

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

  • The same weight-transformation technique could be tested on kernels that are not exactly finite tensor superpositions to see how far the discrete calculus extends.
  • Numerical verification of the embedding constants for small d and specific monotone weights would provide immediate evidence for the infinite-dimensional case.
  • The role of completely monotone weights suggests a possible link to Laplace-transform representations that might simplify computations of the embedding norms.
  • In applications, one could choose the weight transformation to minimize the embedding constant rather than merely to guarantee its finiteness.

Load-bearing premise

The kernels must be superpositions of weighted finite tensor products of one fixed univariate kernel, and the discrete calculus on the weight cone must be sufficient to produce the norm bounds needed for embeddings.

What would settle it

A concrete pair of univariate kernels together with a weight sequence for which no transformed weight sequence satisfies the embedding inequality between the two Hilbert spaces, or a numerical check in infinite dimensions showing that the embedding constant diverges when the weights are not completely monotone.

read the original abstract

We study embeddings between reproducing kernel Hilbert spaces $H(K)$ of functions of $d \in \mathbb{N} \cup \{\infty\}$ variables. The kernels $K$ are superpositions of weighted finite tensor products of a fixed univariate kernel. The basic idea for the embeddings is to compensate a change of the univariate kernel by a suitable transformation of the weights. For the proofs we employ ($d \in \mathbb{N}$) and develop ($d = \infty$) a discrete calculus on the cone of all weights, where completely monotone weights play a particular role. We sketch how to apply the embedding results to computational problems, as, e.g., numerical integration or function recovery.

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 / 3 minor

Summary. The manuscript studies embeddings between reproducing kernel Hilbert spaces H(K) for functions of d variables, where d ranges over the natural numbers and infinity. The kernels are constructed as superpositions of weighted finite tensor products of a fixed univariate kernel. The core technique compensates a change in the univariate kernel by a suitable transformation of the weights. Proofs for finite d employ a discrete calculus on the cone of all weights, while for infinite d this calculus is developed further, with completely monotone weights playing a distinguished role. The embedding results are sketched in application to numerical integration and function recovery.

Significance. If the claimed embeddings and the supporting discrete calculus are rigorously established, the work supplies a systematic method for comparing RKHS norms across different univariate kernels via weight transformations. This extends standard techniques in weighted tensor-product settings to more general weight sequences and to the infinite-dimensional case. The development of a discrete calculus on the weight cone, especially the emphasis on completely monotone weights, is a technical contribution that may yield sharper comparison constants and could be reusable in other high-dimensional approximation problems. The sketched applications indicate relevance to computational tasks such as integration and recovery.

minor comments (3)
  1. [Introduction] The introduction should include a concise statement of the main embedding theorem (including the form of the embedding constant or norm equivalence) to make the central result immediately visible to readers.
  2. [Discrete calculus for infinite d] In the section developing the discrete calculus for d = ∞, the transition from the finite-dimensional case to the infinite case would benefit from an explicit lemma stating the conditions under which the superposition kernels inherit the required monotonicity properties.
  3. [Applications] The application sketches for numerical integration and function recovery remain at a high level; adding one concrete example with an explicit error bound derived from the embedding would strengthen the practical illustration.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary of our work on embeddings of RKHS with general weights and for recognizing the role of the discrete calculus on weight cones, particularly for completely monotone weights in the infinite-dimensional case. We appreciate the recommendation for minor revision.

Circularity Check

0 steps flagged

No significant circularity detected in derivation chain

full rationale

The paper constructs embeddings of RKHS by transforming weights to compensate for changes in a fixed univariate kernel, using a discrete calculus on the weight cone (with emphasis on completely monotone weights) that is explicitly developed within the paper for the infinite-dimensional case and employed for finite d. No step reduces by construction to a fitted input, self-definition, or load-bearing self-citation; the superposition kernel structure is standard, the compensation mechanism follows directly from the introduced monotonicity and ordering tools on weights, and the central claims remain independent of any prior results by the same authors. The derivation is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Based solely on the abstract, the paper relies on standard RKHS axioms and introduces a domain-specific discrete calculus; no free parameters or invented entities are evident at this level.

axioms (2)
  • standard math Kernels are reproducing kernels inducing Hilbert spaces of functions on d variables
    Standard background in functional analysis invoked implicitly for the spaces H(K).
  • domain assumption Weights form a cone admitting a discrete calculus with completely monotone elements having special closure properties
    Central to the embedding proofs as stated in the abstract.

pith-pipeline@v0.9.0 · 5408 in / 1241 out tokens · 55756 ms · 2026-05-07T10:34:31.323090+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

37 extracted references · 37 canonical work pages

  1. [1]

    Adcock, S

    B. Adcock, S. Brugiapaglia, and C. G. Webster. Sparse Polynomial Approximation of High-Dimensional Functions. SIAM, Philadelphia, 2022

    work page 2022

  2. [2]

    M. Aigner. Combinatorial Theory. Springer, Heidelberg, 1979

    work page 1979

  3. [3]

    Aistleitner and J

    C. Aistleitner and J. Dick. Functions of bounded variation, signed measures, and a general K oksma- H lawka inequality. Acta Arith., 167: 0 143--171, 2015

    work page 2015

  4. [4]

    Aronszajn

    N. Aronszajn. Theory of reproducing kernels. Trans. Amer. Math. Soc., 68: 0 337--404, 1950

    work page 1950

  5. [5]

    Baldeaux and M

    J. Baldeaux and M. Gnewuch. Optimal randomized multilevel algorithms for infinite-dimensional integration on function spaces with ANOVA -type decomposition. SIAM J. Numer. Anal., 52: 0 1128--1155, 2014

    work page 2014

  6. [6]

    Cohen and R

    A. Cohen and R. DeVore. Approximation of high-dimensional parametric PDE s. Acta Numerica, 24: 0 1--159, 2015

    work page 2015

  7. [7]

    Dick and M

    J. Dick and M. Gnewuch. Infinite-dimensional integration in weighted H ilbert spaces: anchored decompositions, optimal deterministic algorithms, and higher order convergence. Found. Comput. Math., 14: 0 1027--1077, 2014

    work page 2014

  8. [8]

    Dick and F

    J. Dick and F. Pillichshammer. Digital Nets and Sequences. Cambridge University Press, Cambridge, 2010

    work page 2010

  9. [9]

    J. Dick, I. H. Sloan, X. Wang, and H. Wo \'z niakowski. Good lattice rules in weighted K orobov spaces with general weights. Numer. Math., 103: 0 63--97, 2006

    work page 2006

  10. [10]

    J. Dick, F. Y. Kuo, and I. H. Sloan. High-dimensional integration -- the quasi- M onte C arlo way. Acta Numerica, 22: 0 133--288, 2013

    work page 2013

  11. [11]

    J. Dick, P. Kritzer, and F. Pillichshammer. Lattice Rules. Springer, Cham, 2022

    work page 2022

  12. [12]

    A. D. Gilbert, F. Y. Kuo, and I. H. Sloan. Equivalence between S obolev spaces of first-order dominating mixed smoothness and unanchored ANOVA spaces on ^d . Math. Comp., 91: 0 1837--1869, 2022

    work page 2022

  13. [13]

    M. Gnewuch. Infinite-dimensional integration on weighted H ilbert spaces. Math. Comp., 81: 0 2175--2205, 2012

    work page 2012

  14. [14]

    Gnewuch, S

    M. Gnewuch, S. Mayer, and K. Ritter. On weighted H ilbert spaces and integration of functions of infinitely many variables. J. Complexity, 30: 0 29--47, 2014

    work page 2014

  15. [15]

    Gnewuch, M

    M. Gnewuch, M. Hefter, A. Hinrichs, and K. Ritter. Embeddings of weighted H ilbert spaces and applications to multivariate and infinite-dimensional integration. J.\ Approx.\ Theory, 222: 0 8--39, 2017 a

    work page 2017

  16. [16]

    Gnewuch, M

    M. Gnewuch, M. Hefter, A. Hinrichs, K. Ritter, and G. W. Wasilkowski. Equivalence of weighted anchored and ANOVA spaces of functions with mixed smoothness of order one in L _p . J.\ Complexity, 40: 0 78--99, 2017 b

    work page 2017

  17. [17]

    Gnewuch, M

    M. Gnewuch, M. Hefter, A. Hinrichs, K. Ritter, and G. W. Wasilkowski. Embeddings for infinite-dimensional integration and L _2 -approximation with increasing smoothness. J.\ Complexity, 54: 0 101406, 1--32, 2019

    work page 2019

  18. [18]

    Gnewuch, A

    M. Gnewuch, A. Hinrichs, K. Ritter, and R. R\"u mann. Infinite-dimensional integration and L^2 -approximation on H ermite spaces. J. Approx. Theory, 300: 0 106027, 1--32, 2024a

    work page

  19. [19]

    Gnewuch , P

    M. Gnewuch , P. Kritzer , A. Owen , and Z. Pan . Computable error bounds for quasi- M onte C arlo using points with non-negative local discrepancy. Inf. Inference, 13: 0 iaae021, 2024b

    work page

  20. [20]

    Gnewuch, K

    M. Gnewuch, K. Ritter, and R. R\"u mann. Multi- and infinite-variate integration and L^2 -approximation on H ilbert spaces with G aussian kernels, 2025. arXiv:2412.05368v2, to appear in Math. Comp. (https://doi.org/10.1090/mcom/4191)

    work page doi:10.1090/mcom/4191 2025

  21. [21]

    Hakula, H

    H. Hakula, H. Harbrecht, V. Kaarnioja, F. Y. Kuo, and I. H. Sloan. Uncertainty quantification for random domains using periodic random variables. Numer.\ Math., 156: 0 273--317, 2024

    work page 2024

  22. [22]

    Hefter and K

    M. Hefter and K. Ritter. On embeddings of weighted tensor product H ilbert spaces. J. Complexity, 31: 0 405--423, 2015

    work page 2015

  23. [23]

    Hefter, K

    M. Hefter, K. Ritter, and G. W. Wasilkowski. On equivalence of weighted anchored and ANOVA spaces of functions with mixed smoothness of order one in L _1 or L _ . J. Complexity, 32: 0 1--19, 2016

    work page 2016

  24. [24]

    F. J. Hickernell and X. Wang. The error bounds and tractability of quasi- M onte C arlo algorithms in infinite dimensions. Math. Comp., 71: 0 1641--1661, 2001

    work page 2001

  25. [25]

    F. J. Hickernell, T. M \"u ller-Gronbach, B. Niu, and K. Ritter. Multi-level M onte C arlo algorithms for infinite-dimensional integration on R ^ . J. Complexity, 26: 0 229--254, 2010

    work page 2010

  26. [26]

    Hinrichs and J

    A. Hinrichs and J. Schneider. Equivalence of anchored and ANOVA spaces via interpolation. J. Complexity, 33: 0 190--198, 2016

    work page 2016

  27. [27]

    F. Y. Kuo and D. Nuyens. Application of quasi- M onte C arlo methods to elliptic PDE s with random diffusion coefficients: a survey of analysis and implementation. Found. Comput. Math., 16: 0 1631--1696, 2016

    work page 2016

  28. [28]

    F. Y. Kuo, I. H. Sloan, G. W. Wasilkowski, and H. Wo\'zniakowski. Liberating the dimension. J. Complexity, 26: 0 422--454, 2010

    work page 2010

  29. [29]

    F. Y. Kuo, C. Schwab, and I. H. Sloan. Quasi- M onte C arlo finite element methods for a class of elliptic partial differential equations with random coefficients. SIAM J. Numer. Anal.., 50: 0 3351--3374, 2012

    work page 2012

  30. [30]

    T. G. Kurtz . Point processes and completely monotone set functions. Z.\ Wahrscheinlichkeits\-theorie verw.\ Gebiete, 31: 0 57--67, 1974

    work page 1974

  31. [31]

    Mat\'u s

    F. Mat\'u s . Extreme convex set functions with many nonnegative differences. Discrete Math., 135: 0 177--191, 1994

    work page 1994

  32. [32]

    Novak and H

    E. Novak and H. Wo\'zniakowski. Tractability of M ultivariate P roblems. Vol. 1: L inear I nformation . EMS Tracts in Mathematics. European Mathematical Society (EMS), Z\"urich, 2008

    work page 2008

  33. [33]

    Novak and H

    E. Novak and H. Wo\'zniakowski. Tractability of M ultivariate P roblems. Vol. 2: S tandard I nformation for F unctionals . EMS Tracts in Mathematics. European Mathematical Society (EMS), Z\"urich, 2010

    work page 2010

  34. [34]

    Novak and H

    E. Novak and H. Wo\'zniakowski. Tractability of M ultivariate P roblems. Vol. 3: S tandard I nformation for O perators . EMS Tracts in Mathematics. European Mathematical Society (EMS), Z\"urich, 2012

    work page 2012

  35. [35]

    V. I. Paulsen and M. Raghupathi. An Introduction to the Theory of Reproducing Kernel H ilbert Spaces . Cambridge Univ.\ Press, Cambridge, 2016

    work page 2016

  36. [36]

    Plaskota and G

    L. Plaskota and G. W. Wasilkowski. Tractability of infinite-dimensional integration in the worst case and randomized settings. J. Complexity, 27: 0 505--518, 2011

    work page 2011

  37. [37]

    I. H. Sloan and H. Wo\' z niakowski. When are quasi- M onte C arlo algorithms efficient for high dimensional integrals? J. Complexity, 14: 0 1--33, 1998

    work page 1998