Pith. sign in

REVIEW 1 major objections 4 minor 21 references

Besov regularity of random wavelet series

T0 review · 1 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read This paper proves that a random wavelet series of the form (1.1) almost surely belongs to the Besov space $B^s_{p,q}(\mathbb{R}^d)$ if and only if the parameters satisfy $\beta<-d/p$ and $s+d/2+\alpha<0$, with the boundary case…

desk verdict Strong new results on non-sparse Besov priors, but the Bernoulli-sparse sufficiency theorem has a genuine boundary counterexample at p=∞, q<∞ that must be fixed before acceptance. read the letter →

arxiv 2411.18155 v1 pith:2SMZPYEO submitted 2024-11-27 math.PR math.FA

Andreas Horst , Thomas Jahn , Felix Voigtlaender This is my paper · ORCID
classification math.PRmath.FA MSC 42B3542C4046F2560G6060H50
keywords randomwaveletseriesBesovregularitypriorKarhunen-Loève-typeexpansionsequencespacemomentgeneratingfunctionBernoullisparsityBayesianinverseproblems
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

This paper studies random wavelet series on $\mathbb{R}^d$ whose coefficients are independent copies of a template random variable multiplied by deterministic scale and shift factors. It proves that, under mild moment assumptions on the template, the series almost surely belongs to a Besov space $B^s_{p,q}(\mathbb{R}^d)$ if and only if the two parameter inequalities $\beta<-d/p$ and $s+d/2+\alpha<0$ hold, with the boundary case $s+d/2+\alpha=0$ requiring $\theta<-1/q$ (and with the obvious replacements at $p=\infty$ or $q=\infty$). The same inequalities characterize finiteness of the $r$-th moment of the Besov norm and, for well-concentrated template variables, finiteness of the exponential moment $E[\exp(c\|f\|^r)]$. A parallel condition governs the Bernoulli-sparse version of the series. A sympathetic reader would care because this turns a probabilistic regularity question into a checkable parameter test and closes gaps left by earlier treatments that required Gaussian or bounded coefficients.

What carries the argument

The workhorse is the wavelet characterization of Besov spaces, which identifies $\|f\|_{B^s_{p,q}}$ (up to equivalence) with the weighted mixed $\ell^p/\ell^q$ norm of the wavelet coefficients. For the sufficiency direction, Lemma 3.1 controls the spatial sum appearing in that norm by a single random quantity $\Xi$, defined as the supremum over dyadic shells of averages of $|\xi|^p$; Property A makes the deterministic weights summable, while the moment condition on $X$ makes $\Xi$ almost surely finite. For the necessity direction, truncated variables, the Paley-Zygmund inequality, and Borel-Cantelli arguments force each individual inequality: if any parameter crosses the threshold, independent blocks of coefficients with positive probability are large enough that the Besov norm diverges almost surely. A separate argument (Theorem 4.2) handles the subtle distinction between the coefficient sequence belonging to $b^s_{p,q}$ and the synthesized function belonging to $B^s_{p,q}$.

What would settle it

Take a template $X$ with $P(X\neq 0)>0$ but $E|X|^p=\infty$ (for example, a tail of order $t^{-p}/\log(t)^2$), choose parameters satisfying Property A, and simulate the Besov sequence norm of the coefficient array. The theorem predicts the norm diverges almost surely, while a weaker moment bound would predict possible convergence.

Watch

Extended reading notes

Core claim

The central claim is a complete if-and-only-if description of Besov regularity for the prior (1.1): with i.i.d. coefficients having a sufficiently nice template law, the random tempered distribution $f$ belongs to $B^s_{p,q}(\mathbb{R}^d)$ almost surely exactly when Property A holds, namely $\gamma<-d/p$, $\beta<-d/p$, and either $s+d/2+\alpha<0$ or $s+d/2+\alpha=0$ with $\theta<-1/q$; when $p=\infty$ the strict inequalities on $\gamma,\beta$ become non-strict, and when $q=\infty$ the strict inequality on $\theta$ becomes non-strict. The same conditions are necessary and sufficient for $E[\|f\|^r_{B^s_{p,q}}]<\infty$ and, under an exponential-moment condition on $|X|^{\max\{r,p\}}$, for $E[\exp(c\|f\|^r)]<\infty$ for some $c>0$. For the Bernoulli-sparse series (1.2) an analogous condition, Property A$'$, is established. The paper also proves that the moment assumptions on the template cannot simply be dropped: essential boundedness is needed in the $p=\infty$ borderline cases, and near-sharpness statements show only mild weakening is possible for $p<\infty$.

Load-bearing premise

The whole characterization rests on the template random variable $X$ having enough integrability: a finite moment $E|X|^{p(1+\varepsilon)}$ when $p<\infty$, and essential boundedness when $p=\infty$, together with $P(X\neq 0)>0$ for the necessity direction.

Editorial extensions

If this is right

  • For the Besov prior (1.1), almost-sure membership in $B^s_{p,q}(\mathbb{R}^d)$ is decided by the two inequalities $\beta<-d/p$ and $s+d/2+\alpha<0$, together with the stated boundary and endpoint modifications.
  • Under the same conditions, and with a well-concentrated template, $E[\|f\|^r_{B^s_{p,q}}]$ and $E[\exp(c\|f\|^r_{B^s_{p,q}})]$ are finite for some $c>0$, so the norm's tail is sub-exponential in the regular regime.
  • The Bernoulli-sparse series (1.2) has an analogous characterization, Property A$'$, in which the sparsity parameter $\mu$ shifts the smoothness threshold by $\mu/p$.
  • The moment assumptions on the template are essentially optimal: for $p=\infty$, boundedness cannot be dropped in the borderline cases, and for $p<\infty$ the conditions can only be weakened mildly.
  • All results transfer to weighted Besov spaces with polynomial weights by a parameter shift explained in Remark 2.7.

Reading between the lines

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

  • The characterization is a sharp phase transition: crossing any single parameter over its threshold switches the series from almost-surely smooth to almost-surely not smooth, which should be visible in numerical samples of the prior.
  • Because the exponential-moment finiteness is tied to the same inequalities, Bayesian inverse problems using these priors inherit well-posedness exactly in the regime identified here, a consequence the paper motivates but does not develop.
  • The Bernoulli-sparse threshold suggests a design rule for sparse priors: to keep a target smoothness while thinning coefficients, one must compensate with $\alpha$ or $\mu$; one could test this by estimating Besov norms of truncated samples.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

1 major / 4 minor

Summary. The paper studies almost-sure Besov regularity, moment finiteness, and exponential moment finiteness for random wavelet series on R^d whose coefficients are products of i.i.d. random variables and deterministic scale/shift decays. For the dense prior (1.1), Theorems 3.3--3.5 and 4.1--4.2 give, under mild moment assumptions on the template variable X, necessary and sufficient conditions (Property A) for f ∈ B^s_{p,q}(R^d) a.s., for E[||f||^r] < ∞, and for E[exp(c||f||^r)] < ∞. For the sparse Bernoulli prior (1.2), Theorems 3.8 and 4.3--4.4 claim analogous but only partially matching conditions (Property A'' versus Property A'). Section 5 discusses sharpness of the moment assumptions, including the p = ∞ endpoint. The proofs are detailed and mostly self-contained, relying on wavelet characterizations, Borel--Cantelli arguments, Paley--Zygmund inequalities, and auxiliary maximal lemmas.

Significance. If the main results were correct, the paper would be a valuable reference for Bayesian inverse problems and for random wavelet series, since it treats the full range p,q ∈ (0,∞], distinguishes p = ∞ and q = ∞ cases carefully, and gives two-sided characterizations rather than only sufficient conditions. The dense-prior theorems appear sound and are supported by self-contained proofs, and the sharpness discussion in Section 5 is a genuine strength. However, the sufficiency statement for the Bernoulli prior, Theorem 3.8, contains a load-bearing error at the p = ∞, q < ∞ boundary, so the claimed complete characterization for sparse priors is not established as stated.

major comments (1)
  1. [§3.8 and Definition 2.8(b)] Theorem 3.8 is false in the stated form for p = ∞, q < ∞. In Step 2 of the proof, the authors bound ~η2 ≤ R ||(2^{j(s+d/2+α)})_{j∈N,t∈T_j}||_{ℓq} and claim this is ≲ R because s+d/2+α ≤ 0 by Property A''. For q < ∞ and s+d/2+α = 0, the sequence is identically 1 over infinitely many scales, so its ℓq norm is infinite. A concrete counterexample is d=1, α=0, β=γ=−1, µ=ν=0, p=∞, q=1, s=−1/2, X≡1. Then Property A'' holds (γ≤0, β≤0, s+d/2+α=0), and the coefficients are a_{j,t,m}=(1+|m|/2^j)^{-1}. The b^{-1/2}_{∞,1} norm is Σ_{j≥1} sup_m (1+|m|/2^j)^{-1} = ∞, so f∉B^{-1/2}_{∞,1}(R) and E[||a||^r]=∞, contradicting Theorem 3.8. The sufficiency condition for p=∞ must distinguish q: for q<∞ one needs s+d/2+α<0, while equality is admissible only for q=∞, matching Property A'(c)/(d). This is a load-bearing issue because it invalidates the claimed sufficiency direction for the Bernoulli-sparse prior in a whole parameter regime.
minor comments (4)
  1. [Lemma A.8] The statement assumes µ,ν ∈ (−∞,0], but the proof contains the line 'since µ ≥ 0 and δ+δε+ε ≥ 0'. This should read 'since µ,ν ≤ 0 (so ̺_{j,τ} ≤ 1) and δ+δε−ε ≥ 0'; the proof also writes 'δ+δε+ε' where 'δ+δε−ε' is meant. As written, the sign conventions are internally inconsistent, though the underlying argument can be repaired.
  2. [Section 2.2] There is a typo in the first paragraph: 'He present a brief outline' should be 'We present a brief outline'.
  3. [Lemma 3.2] The sentence 'Let further (ξ_{j,t,m})... by a family of of random variables' contains a duplicated 'of'.
  4. [Proposition 5.2] The statement begins 'Assume that = (a_{j,t,m})...' and is missing the symbol 'a' before the equals sign.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's characterization derives from the stated stochastic model and standard wavelet/Besov facts, with no fitted parameters or self-referential reductions.

full rationale

The paper defines the random wavelet coefficients explicitly as products of a template random variable X, optional Bernoulli masks, and deterministic scale/location weights (Definition 2.5), and then proves both sufficiency (Section 3) and necessity (Section 4) by direct inequalities. The parameters α, β, γ, θ, µ, ν are inputs to the model, not calibrated to any target quantity. The Besov membership conditions (Properties A, A', A'') are derived from convergence of the wavelet coefficient norm in b_{p,q}^s, using the external wavelet characterization Theorem 2.3 (from Triebel) and standard probabilistic tools such as Kallenberg's Lemma A.2 and the Paley–Zygmund inequality. No theorem of the paper is assumed to prove itself; the sufficiency proofs bound the random norm by auxiliary quantities Ξ or ~Ξ whose finiteness is established independently from the moment assumptions, and the necessity proofs use Borel–Cantelli and Paley–Zygmund to force the parameter inequalities. The one author self-citation, Theorem 2.2 taken from [12] (Grohs–Klotz–Voigtlaender), is a standard existence statement for compactly supported wavelets and is not the paper's central claim; it is an external input, not a result that presupposes the Besov regularity characterization. The reviewer's expressed concern about the p = ∞, q < ∞ boundary in Theorem 3.8 is a potential mathematical error in a specific estimate, not a circular dependence on the paper's own inputs, and therefore does not affect the circularity score. Overall, there is no circular step, no fitted-input-as-prediction, and no self-citation chain bearing the main result.

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

The central theorem takes the stochastic model as input; no constants are fitted, and the only substantive inputs are standard wavelet characterization theorems and classical probabilistic lemmas. The moment condition on X is a model assumption shown to be near sharp in Section 5.

assumptions (5)
  • standard math There exists an orthonormal wavelet basis of L2(R^d) with compactly supported C^k scaling and wavelet functions and k vanishing moments (Theorem 2.2, following [12, Theorem 10]).
    This is the basis used to define the Besov prior and to invoke the wavelet characterization of Besov spaces.
  • standard math The wavelet characterization of Besov spaces: the synthesis operator S maps b^s_{p,q}(R^d) isomorphically onto B^s_{p,q}(R^d) for wavelets of sufficient smoothness (Theorem 2.3, following [21, Theorem 1.26]).
    This bridges the coefficient-sequence norm and the function norm and is used in both the sufficiency and necessity arguments.
  • standard math Classical probabilistic tools: Borel-Cantelli lemmas, the Paley-Zygmund inequality (Lemma A.4), and the zero-one criterion for almost sure finiteness of sums of independent nonnegative random variables (Lemma A.2).
    These tools convert almost sure finiteness of the Besov norm into summability conditions on the model parameters in the necessity proofs.
  • domain assumption The random coefficients xi_{j,t,m} are i.i.d. copies of the template variable X, independent of the Bernoulli variables in the sparse model, with P(X != 0) > 0 and, for p < infinity, E[|X|^{p(1+epsilon)}] < infinity, or essential boundedness for p = infinity.
    This is the stochastic model from Definition 2.5. Section 5 proves that these moment conditions are near sharp, so they are load-bearing rather than cosmetic.
  • standard math The weight w_sigma(x) = (1 + ||x||^2)^{sigma/2} is admissible in the sense of Definition 2.1 (Lemma A.1).
    Used in Theorems 4.2 and 4.4 to transfer regularity from a weighted coefficient space back to the unweighted sequence space.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Besov regularity of random wavelet series." pith.science (2026). https://pith.science/paper/2SMZPYEO

@misc{pith2026241118155,
  author       = {Pith},
  title        = {Pith review of: Besov regularity of random wavelet series},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2SMZPYEO}},
  note         = {Machine review of arXiv:2411.18155}
}
abstract

We study the Besov regularity of wavelet series on $\mathbb{R}^d$ with randomly chosen coefficients. More precisely, each coefficient is a product of a random factor and a parameterized deterministic factor (decaying with the scale $j$ and the norm of the shift $m$). Compared to the literature, we impose relatively mild conditions on the moments of the random variables in order to characterize the almost sure convergence of the wavelet series in Besov spaces $B^s_{p,q}(\mathbb{R}^d)$ and the finiteness of the moments as well as of the moment generating function of the Besov norm. In most cases, we achieve a complete characterization, i.e., the derived conditions are both necessary and sufficient.

Figures

Figures reproduced from arXiv: 2411.18155 by the authors.

Figure 1
Figure 1. Samples of our Besov prior for different sets of param￾eters. The parameters β and γ impact the spatial decay of the function, whereas the smoothness changes with α. In all panels, we set θ = 0. Remark 2.7. In the sequel, we will work with unweighted Besov spaces Bs p,q(Rd ) and b s p,q(Rd ). However, the entire analysis can be readily transferred to weighted Besov spaces with weight function wσ : Rd → R, wσ(x) := (… view at source ↗
Figure 2
Figure 2. Samples of our Besov–Bernoulli prior for different sets of parameters. The parameter ν impacts the spatial decay, whereas the smoothness changes with µ. In all panels, we set α = γ = β = −0.5. where g follows a Besov prior with parameters (α, β + σ, γ + σ, θ, k) and template random variable X. This holds true under the assumption that (2.7) is finite. The terminology introduced in the following definitions will grea… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 20 canonical work pages

  1. [1]

    Abramovich, T

    F. Abramovich, T. Sapatinas, and B.W. Silverman. W avele t thresholding via a Bayesian approach. J. R. Stat. Soc. Ser. B Stat. Methodol. , 60(4):725–749, 1998

    work page 1998

  2. [2]

    Abramowitz and I.A

    M. Abramowitz and I.A. Stegun. Handbook of mathematical functions with formulas, graphs, and mathematical tables , volume No. 55 of National Bureau of Standards Applied Mathemat- ics Series . U. S. Government Printing Office, W ashington, DC, 1964

    work page 1964

  3. [3]

    Aubry and S

    J.-M. Aubry and S. Jaffard. Random wavelet series. Comm. Math. Phys. , 227(3):483–514, 2002

    work page 2002

  4. [4]

    Aziznejad and J

    S. Aziznejad and J. Fageot. W avelet analysis of the Besov regularity of Lévy white noise. Electron. J. Probab. , 25:Paper No. 158, 38, 2020

    work page 2020

  5. [5]

    H. Bauer. Probability theory, volume 23 of De Gruyter Stud. Math. Berlin: de Gruyter, 1996

    work page 1996

  6. [6]

    Besov regularity of functions with sparse random wavelet coefficients

    Natalia Bochkina. Besov regularity of functions with sp arse random wavelet coefficients. arXiv preprint arXiv:1310.3720, 2013. 50 ANDREAS HORST, THOMAS JAHN, AND FELIX VOIGTLAENDER

    work page Pith review arXiv 2013

  7. [7]

    Cioica, S

    P.A. Cioica, S. Dahlke, N. Döhring, S. Kinzel, F. Lindner , T. Raasch, K. Ritter, and R.L. Schilling. Adaptive wavelet methods for SPDEs. In Extraction of quantifiable information from complex systems , volume 102 of Lect. Notes Comput. Sci. Eng. , pages 83–107. Springer, Cham, 2014

    work page 2014

  8. [8]

    Dashti, S

    M. Dashti, S. Harris, and A. Stuart. Besov priors for Baye sian inverse problems. Inverse Probl. Imaging , 6(2):183–200, 2012

    work page 2012

Show all 21 references
  1. [9]

    Daubechies

    I. Daubechies. Ten lectures on wavelets, volume 61 of CBMS-NSF Regional Conference Series in Applied Mathematics . Society for Industrial and Applied Mathematics (SIAM), Ph iladel- phia, PA, 1992

    1992

  2. [10]

    Esser, S

    C. Esser, S. Jaffard, and B. Vedel. Regularity propertie s of random wavelet series. Theory Probab. Math. Statist. , 110:31–53, 2024

    2024

  3. [11]

    G.B. Folland. Real analysis. Pure and Applied Mathematics (New York). John Wiley & Sons, Inc., New York, second edition, 1999. Modern techniques and their applications, A Wiley- Interscience Publication

    1999

  4. [12]

    Grohs, A

    P. Grohs, A. Klotz, and F. Voigtlaender. Phase transiti ons in rate distortion theory and deep learning. Found. Comput. Math. , 23(1):329–392, 2023

    2023

  5. [13]

    Haroske and H

    D. Haroske and H. Triebel. Entropy numbers in weighted f unction spaces and eigenvalue distributions of some degenerate pseudodifferential opera tors. I. Math. Nachr. , 167:131–156, 1994

    1994

  6. [14]

    Kallenberg

    O. Kallenberg. Foundations of modern probability , volume 99 of Probability Theory and Sto- chastic Modelling . Springer, Cham, 3rd edition, 2021

    2021

  7. [15]

    Kekkonen, M

    H. Kekkonen, M. Lassas, E. Saksman, and S. Siltanen. Ran dom tree Besov priors—towards fractal imaging. Inverse Probl. Imaging , 17(2):507–531, 2023

    2023

  8. [16]

    Knoblauch

    A. Knoblauch. Closed-form expressions for the moments of the binomial probability distri- bution. SIAM J. Appl. Math. , 69(1):197–204, 2008

    2008

  9. [17]

    S. Lan, M. Pasha, S. Li, and W. Shen. Spatiotemporal Beso v priors for bayesian inverse problems. arXiv preprint arXiv:2306.16378 , 2023

    2023

  10. [18]

    Lassas, E

    M. Lassas, E. Saksman, and S. Siltanen. Discretization -invariant Bayesian inversion and Besov space priors. Inverse Probl. Imaging , 3(1):87–122, 2009

    2009

  11. [19]

    H. Triebel. Higher analysis . Hochschulbücher für Mathematik. [University Books for Ma the- matics]. Johann Ambrosius Barth Verlag GmbH, Leipzig, 1992

    1992

  12. [20]

    H. Triebel. Theory of function spaces. III , volume 100 of Monographs in Mathematics . Birkhäuser Verlag, Basel, 2006

    2006

  13. [21]

    H. Triebel. Function spaces and wavelets on domains , volume 7 of EMS Tracts in Mathe- matics. European Mathematical Society (EMS), Zürich, 2008. (A. Horst) Technical University of Denmark, Department of Applied Math ematics and Computer Science, Matematiktor vet 303B, 2800 Kg...

    2008

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.