Pith. sign in

REVIEW 6 minor 1 cited by

Sharp uniform approximation for spectral Barron functions by deep neural networks

T0 review · 0 major / 6 minor · reviewed 2026-08-06 · 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 spectral Barron functions of smoothness $s$ are approximated by deep ReLU networks in $L^p$ at the sharp rate $N^{-sL}$ whenever $0<sL\le 1/2$.

desk verdict Solid extension of the L2 theory to Lp and uniform rates for spectral Barron functions; the main theorems hold, and only the abstract wording and omitted corollary details need fixing. read the letter →

arxiv 2507.06789 v1 pith:FJKKCFCS submitted 2025-07-09 math.NA cs.NA

Yulei Liao , Pingbing Ming , Hao Yu This is my paper · ORCID
classification math.NAcs.NA MSC 41A2541A4642A3868T07
keywords spectralBarronspacedeepneuralnetworksReLULpapproximationuniformsharpratesMonteCarloratedimension-freeconstants
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

Approximation theory for neural networks traditionally required spectral Barron functions of smoothness at least $1$ to attain the Monte Carlo rate $N^{-1/2}$. This paper shows that smoothness $1/2$ is enough for shallow networks, and that with $L$ hidden layers any smoothness $s$ with $0

What carries the argument

The engine is the integral representation $f(x)=\pi^2\int_{\mathbb{R}^d}\int_0^1|\widehat f(\xi)|\sin(\pi r)\gamma_{n_\xi}(t_\xi(x),r)\,dr\,d\xi$, in which $\gamma(t,r)$ is a ReLU-computable hat-shaped wavelet symmetric about $1/2$, and its repeated-period version $\gamma_n(\cdot,r)$ realizes $\cos(2\pi n t)$ through $\pi^2\int_0^1\sin(\pi r)\gamma_n(t,r)\,dr=\cos(2\pi n t)$. The composition identity $\gamma_{n_2}\circ\beta_{n_1}=\gamma_{2n_1n_2}$ (Lemma 3.2, from the companion paper [20]) lets each hidden layer multiply frequency, so choosing $n_i\approx(1+|\xi|_1)^{1/L}$ keeps the per-layer width at $4\lceil(1+|\xi|_1)^{1/L}\rceil$ and converts a Monte Carlo sample of size $N$ into the global rate $N^{-sL}$. The $L^\infty$ proof adds a Rademacher-complexity step with a metric-entropy bound on the class $x\mapsto\gamma_{n_\xi}(t_\xi(x),r)$, while the lower bound counts sign changes of $\cos(2\pi n x_1)$ to show that any $N$-unit-per-layer network must miss oscillation intervals.

What would settle it

Evaluate the identity $\gamma_{n_2}\circ\beta_{n_1}=\gamma_{2n_1n_2}$ at the breakpoints $t=j/(2n_1n_2)$, $j=0,\dots,2n_1n_2$, for small parameters such as $n_1=2$, $n_2=3$; a nonzero difference at any of those points would invalidate the width bound behind Theorem 3.1 and the claimed $N^{-sL}$ rate.

Watch

Extended reading notes

Core claim

The central discovery is that the approximation rate for the spectral Barron space $\mathscr{B}^s(\Omega)$ is exactly $N^{-sL}$ in $L^p$ whenever the depth-smoothness product satisfies $0<sL\le 1/2$, with constants independent of the dimension. Theorem 3.1 constructs an $(L,N)$-ReLU network with $\|f-f_{L,N}\|_{L^p(\Omega)}\le 13\sqrt{p}\|f\|_{\mathscr{B}^s(\Omega)}N^{-sL}$ for $p\ge2$; Theorem 3.6 gives the uniform analogue $45\|f\|_{\mathscr{B}^s(\Omega)}N^{-sL}\sqrt{1+dL\ln N}$; and Theorem 3.10 proves a matching lower bound of order $N^{-sL}$ for a worst-case function with $\|f\|_{\mathscr{B}^s(\Omega)}\le 2+\varepsilon$. Shallow sigmoidal networks already attain $N^{-s}$ for $0<s\le1/2$, improving the earlier smoothness requirement $s\ge1$ for the Monte Carlo rate and the earlier deep rate $N^{-sL/2}$ to the sharp exponent $sL$.

Load-bearing premise

Everything in the deep part rests on an exact frequency-doubling identity for the paper's repeated wave shapes: composing a wave with a triangle wave must reproduce the wave with twice as many periods on the whole interval $[0,1]$, including at the breakpoints; if that identity fails at a single point, the per-layer width bound and the $N^{-sL}$ rate no longer follow.

Editorial extensions

If this is right

  • Functions in $\mathscr{B}^s$ with $0<sL\le1/2$, including functions too rough to be Hölder continuous of order above $s$, are approximated at order $N^{-sL}$ by deep ReLU networks with dimension-free prefactors.
  • Shallow networks with one hidden layer reach the Monte Carlo rate $N^{-1/2}$ in $L^p$ for $\mathscr{B}^{1/2}$ functions, and the same rate in $L^\infty$ up to a logarithmic factor.
  • The matching lower bound means the exponent $sL$ is optimal: no ReLU or Heaviside network with $L$ hidden layers and $N$ units per layer can approximate all of $\mathscr{B}^s$ at a better worst-case order in this range.
  • Increasing depth directly improves the order for small smoothness: with $s=0.05$, the paper notes that 2, 6, and 10 hidden layers give rates approaching $N^{-1/10}$, $N^{-3/10}$, and $N^{-1/2}$, respectively.
  • The uniform bound loses only a $\sqrt{dL\ln N}$ factor, so the dimension-free character of the Monte Carlo rate survives in the sup norm.

Reading between the lines

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

  • A testable extension of the same machinery: construct the paper's benchmark $f(x)=n^{-s}\cos(2\pi n x_1)e^{-\pi|x|^2/R}$ and numerically measure the best $L^p$ error of $(L,N)$-ReLU networks; the predicted exponent $sL$ should appear for every $0<sL\le1/2$, and any systematic slowing would point at the composition identity rather than the Monte Carlo step.
  • The paper's observation that sigmoidal activations reduce to Heaviside by shifting and scaling suggests the deep rates may transfer to tanh, softplus, ELU, and ReLU$^k$ networks; this transfer is not proved for depth here and is the natural next check.
  • Since the $L^\infty$ upper bound carries a $\sqrt{\ln N}$ factor and the $L^p$ bounds do not, the open question left implicit is whether the sup-norm logarithmic factor is removable; that would need a matching $L^\infty$ lower bound at logarithmically finer scales.
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

0 major / 6 minor

Summary. This paper proves non-asymptotic approximation bounds for spectral Barron functions by shallow sigmoidal and deep ReLU networks. For f in B^s(Omega) with 0 < s <= 1/2 and p >= 2, Theorem 2.4 gives a shallow sigmoidal network with N units satisfying ||f - f_N||_{Lp} <= 28 pi sqrt(p) ||f||_{B^s} N^{-s}. Theorem 3.1 extends the sharp rate N^{-sL} to (L,N)-ReLU networks under 0 < sL <= 1/2, with constant 13 sqrt(p). Theorem 3.6 establishes the uniform bound ||f - f_{L,N}||_{Linfty} <= 45 ||f||_{B^s} N^{-sL} sqrt(1 + dL ln N). Theorem 3.10 provides a matching L^p lower bound for 1 <= p <= infinity, showing the rates are sharp up to a logarithmic factor. Corollaries refine the bounds to depend on the seminorm upsilon_{f,s}; Appendix A proves a sharp bound on the Khintchine constant, and Appendix B shows the embedding B^s into C^s is essentially sharp.

Significance. The results are significant because they lower the smoothness threshold for Monte Carlo rates in L^p from B^1 to B^{1/2} and, more generally, show that deeper networks improve the rate for rough spectral Barron functions. The proofs are substantial: they combine multiscale cosine expansions, symmetrization with Khintchine inequalities, Rademacher complexity, and metric-entropy estimates, and the constants are explicit. The lower bound by oscillation counting has explicit constants and confirms sharpness. I found no error that touches the central claims; I also directly checked the composition identity in Lemma 3.2, and it holds, so the reader's flagged weakest assumption does not land. The main caveat is the paper's reliance on the companion paper [20] for Lemma 3.2 and Lemma 3.11, which should be verified as accepted and publicly available.

minor comments (6)
  1. [Abstract; Theorem 3.6; Corollary 3.9] The abstract states that 'the rates and prefactors in our estimates are dimension-free', but Theorem 3.6 and Corollary 3.9 have prefactors sqrt(1 + dL ln N) and (1 + dL ln N)^{sL}, respectively; please qualify the claim, for example by saying 'dimension-free up to logarithmic factors in d'.
  2. [Section 3, Corollaries 3.4 and 3.9; final L^p improvement] Corollaries 3.4 and 3.9, as well as the final unnumbered L^p improvement in Section 3.2, are stated without proofs ('details omitted for brevity'); since these corollaries are advertised results, please provide the full arguments or a detailed proof sketch with the relevant norm tracking.
  3. [Proof of Theorem 2.4, bound on ||F(.,xi,r)||_{Linfty(Omega)}] With |alpha_{xi,l,j}| <= 2^{1-l} pi and the prefactor 2^{(1+s)l}, the displayed constant 2^{1+sr} in the bound on ||F(.,xi,r)||_{Linfty(Omega)} appears to be missing a factor 2; the final rate is unaffected, but the estimate should be corrected.
  4. [Theorem 3.6, after Eq. (3.12)] The optimization of \tilde c and the conclusion 'Setting \tilde c = 16.26^{-1} ... since c0 < 1' is too compressed; please include the intermediate inequalities that justify the final logarithmic factor, and fix the notation 16.26^{-1} to read 16.26^{-1}.
  5. [Lemma 3.2 and Lemma 3.11] Lemma 3.2 and Lemma 3.11 are quoted from the companion paper [20]; because Lemma 3.2 is the key step that turns depth into the N^{-sL} rate, please either reproduce the proof or give a precise citation to the lemma's proof in [20].
  6. [General presentation] There are several typographical and formatting issues, including 'n-internals' in the proof of Theorem 3.10 (should be 'n intervals') and the rendering of 'H\"older' in a few places; these should be cleaned up before publication.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; the central Lp and L∞ theorems are proven directly from Fourier representations and probabilistic estimates, with self-cited lemmas serving as parameter-free technical support.

full rationale

The main upper bounds (Theorems 3.1 and 3.6) are derived in the paper from the integral representation of Barron functions, the hat-function kernels beta and gamma, symmetrization, the Khintchine inequality, and metric-entropy bounds for the Rademacher complexity. They do not invoke the authors' earlier L2 rate as an input; the L2 case is recovered as a special case of Corollary 2.9 rather than used as a premise. Lemma 3.2, quoted from the authors' companion paper [20], is load-bearing for the depth-composition argument, but it is an elementary parameter-free identity (g_{n2} composed with beta_{n1} equals g_{2n1n2}) whose stated assumptions do not include any target rate, and the paper supplies the definitions and symmetry verification needed to make it checkable; this is ordinary self-citation, not circularity under the review rules. Lemma 3.11 is likewise a parameter-free Fourier estimate used only to place the lower-bound example in B^s, not a restatement of the approximation bound. The lower bound Theorem 3.10 uses standard oscillation counting from Telgarsky together with this independent spectral bound. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and no known result is merely repackaged under new coordinates. The omissions of routine details in Corollaries 3.4, 3.9, and the final Lp improvement are expositional completeness issues, not circular reductions. The abstract's 'dimension-free prefactors' versus the sqrt(1+dL ln N) factor in Theorem 3.6 is a consistency concern, not a circularity concern.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The paper introduces no new entities and no fitted parameters. It depends on the spectral Barron space definition, standard probabilistic inequalities, and two specialized lemmas from the authors' companion paper [20] (the composition identity and the spectral bound for Gaussian-windowed cosines). These are the main upstream costs of the proof.

assumptions (6)
  • domain assumption Fourier inversion and weighted L1 integrability for f in B^s(Omega), including existence of an extension attaining the infimum.
    Used in the proofs of Theorems 2.4 and 3.1 ('we continue to use f to denote the extension... that attains the infimum'); this is the definition of B^s and [20, Lemma 2.1].
  • standard math Khintchine inequality with optimal constants and the bound C_p <= sqrt(p/2) (Lemmas 2.7 and 2.8).
    Core to the Lp Monte Carlo estimates in (2.7)-(2.8) and in Theorem 3.1; Haagerup's theorem is cited as [16].
  • standard math Composition identity gamma_{n2} composed with beta_{n1} equals gamma_{2 n1 n2} on [0,1] (Lemma 3.2, cited from [20, Lemma 3.5]).
    This exact identity determines the per-layer width 4 ceil((1+|xi|)^{1/L}) and is the load-bearing deep network encoding; it is not reproved in this paper.
  • standard math Gaussian-windowed cosine spectral bound: upsilon_{f,s} <= (n + d/(pi sqrt(R)))^s for f = cos(2 pi n x1) e^{-pi |x|^2/R}, 0<=s<=1 (Lemma 3.11, from [20, Lemma SM6.1]).
    Used in lower bound Theorem 3.10 to construct a B^s function with controlled norm and many oscillations.
  • standard math Telgarsky's linear-region and sign-change count for L-hidden ReLU networks (cited as [39, Lemma 3.2]).
    Used in Theorem 3.10 to guarantee many intervals where the approximant does not change sign.
  • domain assumption The phase shift theta(xi) can be chosen so that 0 <= xi dot x + theta(xi) <= 1 + |xi|_1 for all x in Omega.
    Invoked in the proof of Theorem 3.1 before defining t_xi(x); it holds because the range of xi dot x over Omega has length at most |xi|_1 and theta is defined modulo integers.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Sharp uniform approximation for spectral Barron functions by deep neural networks." pith.science (2026). https://pith.science/paper/FJKKCFCS

@misc{pith2026250706789,
  author       = {Pith},
  title        = {Pith review of: Sharp uniform approximation for spectral Barron functions by deep neural networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FJKKCFCS}},
  note         = {Machine review of arXiv:2507.06789}
}
abstract

This work explores the neural network approximation capabilities for functions within the spectral Barron space $\mathscr{B}^s$, where $s$ is the smoothness index. We demonstrate that for functions in $\mathscr{B}^{1/2}$, a shallow neural network (a single hidden layer) with $N$ units can achieve an $L^p$-approximation rate of $\mathcal{O}(N^{-1/2})$. This rate also applies to uniform approximation, differing by at most a logarithmic factor. Our results significantly reduce the smoothness requirement compared to existing theory, which necessitate functions to belong to $\mathscr{B}^1$ in order to attain the same rate. Furthermore, we show that increasing the network's depth can notably improve the approximation order for functions with small smoothness. Specifically, for networks with $L$ hidden layers, functions in $\mathscr{B}^s$ with $0 < sL \le 1/2$ can achieve an approximation rate of $\mathcal{O}(N^{-sL})$. The rates and prefactors in our estimates are dimension-free. We also confirm the sharpness of our findings, with the lower bound closely aligning with the upper, with a discrepancy of at most one logarithmic factor.

Figures

Figures reproduced from arXiv: 2507.06789 by the authors.

Figure 1
Figure 1. The plots of f defined in (B.1). Left: d = 1 with different α. Right: d = 2 and α = 0.05. By (B.2), f defined in (B.1) is a radial function with the representation f(x) = h 2 d+α/2−1π d/2Γ((α + d)/2)i−1 |x| α/2Kα/2(|x|), where Kν is modified Bessel function of the second kind Kν(z) = Γ(ν + 1/2)(2z) ν √ π Z ∞ 0 [PITH_FULL_IMAGE:figures/full_fig_p025_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Barron regularity of many particle Schr\"odinger eigenfunctions

    math.AP 2025-08 accept novelty 7.0 of 10

    Many-particle Schrödinger eigenfunctions with singular potentials are shown to lie in spectral Barron spaces up to a sharp smoothness index, giving the missing regularity theory for neural-network quantum solvers.

Reference graph

Works this paper leans on

44 extracted references · 43 canonical work pages · cited by 1 Pith paper

  1. [20]

    Liao and P

    Y. Liao and P. Ming, Spectral Barron space for deep neural network approximatio n, SIAM J. Math. Data Sci. (accepted). arXiv:2309.00788. SHARP UNIFORM APPROX. FOR SPECTRAL BARRON FUNCTIONS BY DNN 2 7

    arXiv

  2. [9]

    , Sharp representation theorems for ReLU networks with preci se dependence on depth , Advances in Neural Information Processing Systems 33 (2020), 10697–10706

    work page 2020

  3. [1]

    E. M. Achour, A. Foucault, S. Gerchinovitz, and F. Malgou yres, A general approximation lower bound in Lp norm, with applications to feed-forward neural networks , Advances in Neural Information Processing Systems 35 (2022), 22396–22408

    work page 2022

  4. [2]

    Artin, The Gamma function , Athena Series: Selected Topics in Mathematics, Holt, Rine - hart and Winston, 1964

    E. Artin, The Gamma function , Athena Series: Selected Topics in Mathematics, Holt, Rine - hart and Winston, 1964

    work page 1964

  5. [3]

    Bach, Breaking the curse of dimensionality with convex neutral ne tworks, J

    F. Bach, Breaking the curse of dimensionality with convex neutral ne tworks, J. Mach. Learn. Res. 18 (2017)

    work page 2017

  6. [4]

    Ball, Cube slicing in Rn, Proc

    K. Ball, Cube slicing in Rn, Proc. Amer. Math. Soc. 97 (1986), no. 3, 465–473

    work page 1986

  7. [5]

    A. R. Barron, Neural net approximation , Proc. 7th Yale W orkshop on Adaptive and Learning Systems 1 (1992), 69–72. [6] , Universal approximation bounds for superpositions of a sig moidal function , IEEE Trans. Inform. Theory 39 (1993), no. 3, 930–945

    work page 1992

  8. [7]

    , Approximation and estimation bounds for artificial neural n etworks, Mach. Learn. 14 (1994), 115–133

    work page 1994

Show all 44 references
  1. [8]

    Bresler and D

    G. Bresler and D. Nagaraj, A corrective view of neural networks: Representation, memo riza- tion and learning , Proceedings of Machine Learning Research 125 (2020), 848–901

    2020

  2. [10]

    Caragea, P

    A. Caragea, P. Petersen, and F. Voigtlaender, Neural network approximation and estimation of classifiers with classification boundary in a Barron class , Ann. Appl. Probab. 33 (2023), no. 4, 3039–3079

    2023

  3. [11]

    DeVore, B

    R. DeVore, B. Hanin, and G. Petrova, Neural network approximation, Acta Numer. 30 (2021), 327–444

    2021

  4. [12]

    M. J. Donahue, L. Gurvits, C. Darken, and E. Sontag, Rates of convex approximation in non-Hilbert spaces, Constr. Approx. 13 (1997), no. 2, 187–220. MR1437210

    1997

  5. [13]

    W. E and Q. W ang, Exponential convergence of the deep neural network approxi mation for analytic functions , Sci. China Math. 61 (2018), no. 10, 1733–1740

    2018

  6. [14]

    Elbr¨ achter, D

    D. Elbr¨ achter, D. Perekrestenko, P. Grohs, and H. B¨ ol cskei, Deep neural network approxi- mation theory , IEEE Trans. Inform. Theory 67 (2021), no. 5, 2581–2623

    2021

  7. [15]

    Eldan and O

    R. Eldan and O. Shamir, The power of depth for feedforward neural networks , Proceedings of Machine Learning Research 49 (2016), 907–940

    2016

  8. [16]

    Haagerup, The best constants in the Khintchine inequality , Studia Math

    U. Haagerup, The best constants in the Khintchine inequality , Studia Math. 70 (1982), no. 3, 231–283

    1982

  9. [17]

    H¨ ormander, Linear partial differential operators , Grundlehren der mathematischen Wis- senschaften, vol

    L. H¨ ormander, Linear partial differential operators , Grundlehren der mathematischen Wis- senschaften, vol. 116, Springer Berlin, Heidelberg, 1963

    1963

  10. [18]

    Y. Jiao, Y. Lai, X. Lu, F. W ang, J. Z. Yang, and Y. Yang, Deep neural networks with ReLU- sine-exponential activations break curse of dimensionali ty in approximation on h¨ older class, SIAM J. Math. Anal. 55 (2023), no. 4, 3635–3649

    2023

  11. [19]

    J. M. Klusowski and A. R. Barron, Approximation by combinations of ReLU and squared ReLU ridge functions with ℓ1 and ℓ0 controls, IEEE Trans. Inform. Theory 64 (2018), no. 12, 7649–7656

    2018

  12. [21]

    J. Lu, Z. Shen, H. Yang, and S. Zhang, Deep network approximation for smooth functions , SIAM J. Math. Anal. 53 (2021), no. 5, 5465–5506

    2021

  13. [22]

    Y. L. Luke, Integrals of Bessel Functions , McGraw-Hill Book Company, 1962

    1962

  14. [23]

    L. Ma, J. W. Siegel, and J. Xu, Uniform approximation rates and metric entropy of shallow neural networks , Res. Math. Sci. 9 (2022), no. 3

    2022

  15. [24]

    Makovoz, Random approximants and neural networks , J

    Y. Makovoz, Random approximants and neural networks , J. Approx. Theory 85 (1996), no. 1, 98–109

    1996

  16. [25]

    Meng and P

    Y. Meng and P. Ming, A new function space from Barron class and application to neu ral network approximation, Commun. Comput. Phys. 32 (2022), no. 5, 1361–1400

    2022

  17. [26]

    V. A. Mikhailets and A. A. Murach, H¨ ormander Spaces, Interpolation, and Elliptic Problems, De Gruyter Studies in Mathematics, vol. 60, W alter de Gruyte r GmbH, Berlin/Boston, 2014

    2014

  18. [27]

    Montanelli and Q

    H. Montanelli and Q. Du, New error bounds for deep ReLU networks using sparse grids , SIAM J. Math. Data Sci. 1 (2019), no. 1, 78–92

    2019

  19. [28]

    Montanelli, H

    H. Montanelli, H. Yang, and Q. Du, Deep ReLU networks overcome the curse of dimension- ality for generalized bandlimited functions , J. Comput. Math. 39 (2021), no. 6, 801–815

    2021

  20. [29]

    Pilipovi´ c, N

    S. Pilipovi´ c, N. Teofanov, and J. Toft, Micro-local analysis in Fourier Lebesgue and modu- lation spaces: part II , J. Pseudo-Differ. Oper. Appl. 1 (2010), no. 3, 341–376

    2010

  21. [30]

    Pinkus, Approximation theory of the MLP model in neural networks , Acta Numer

    A. Pinkus, Approximation theory of the MLP model in neural networks , Acta Numer. 8 (1999), 143–195

    1999

  22. [31]

    Rebeschini, Lecture notes in algorithmic foundations of learning: Cove ring numbers bounds for Rademacher complexity

    P. Rebeschini, Lecture notes in algorithmic foundations of learning: Cove ring numbers bounds for Rademacher complexity. chaining , 2020

    2020

  23. [32]

    Safran, R

    I. Safran, R. Eldan, and O. Shamir, Depth separations in neural networks: What is actually being separated?, Constr. Approx. 55 (2022), no. 1, 225–257

    2022

  24. [33]

    Z. Shen, H. Yang, and S. Zhang, Optimal approximation rate of ReLU networks in terms of width and depth , J. Math. Pures Appl. (9) 157 (2022), 101–135

    2022

  25. [34]

    J. W. Siegel, Optimal approximation of zonoids and uniform approximatio n by shallow neural networks, 2025. arXiv:2307.15285

    2025 arXiv

  26. [35]

    J. W. Siegel and J. Xu, Approximation rates for neural networks with general activ ation functions, Neural Networks 128 (2020), 313–321

    2020

  27. [36]

    , High-order approximation rates for shallow neural network s with cosine and ReLU k activation functions , Appl. Comput. Harmon. Anal. 58 (2022), 1–26

    2022

  28. [37]

    , Sharp bounds on the approximation rates, metric entropy, an d n-widths of shallow neural networks , Found. Comput. Math. 24 (2024), 481–537

    2024

  29. [38]

    E. M. Stein and G. W eiss, Introduction to Fourier Analysis on Euclidean Spaces , Princeton Mathematical Series, Princeton University Press, 1971

    1971

  30. [39]

    Telgarsky, Benefits of depth in neural networks , Proceedings of Machine Learning Re- search 49 (2016), 1517–1539

    M. Telgarsky, Benefits of depth in neural networks , Proceedings of Machine Learning Re- search 49 (2016), 1517–1539

    2016

  31. [40]

    Tsuji and T

    K. Tsuji and T. Suzuki, Estimation error analysis of deep learning on the regressio n problem on the variable exponent Besov space , Electron. J. Stat. 15 (2021), no. 1, 1869–1908

    2021

  32. [41]

    Xu, Finite neuron method and convergence analysis , Commun

    J. Xu, Finite neuron method and convergence analysis , Commun. Comput. Phys. 28 (2020), no. 5, 1707–1745

    2020

  33. [42]

    Yarotsky, Error bounds for approximations with deep ReLU networks , Neural networks 94 (2017), 103–114

    D. Yarotsky, Error bounds for approximations with deep ReLU networks , Neural networks 94 (2017), 103–114

    2017

  34. [43]

    , Optimal approximation of continuous functions by very deep ReLU networks , Pro- ceedings of the 31st conference on learning theory, 2018, pp . 639–649

    2018

  35. [44]

    Yarotsky and A

    D. Yarotsky and A. Zhevnerchuk, The phase diagram of approximation rates for deep neural networks, Advances in neural information processing systems, 2020, pp. 13005–13015. 28 Y. L. LIAO, P. B. MING, AND H. YU

    2020

  36. [45]

    J. E. Yukich, M. B. Stinchcombe, and H. White, Sup-norm approximation bounds for net- works through probabilistic methods , IEEE Trans. Inform. Theory 41 (1995), no. 4, 1021– 1027. Department of Mathematics, F aculty of Science, National Univ ersity of Singapore, 10 Lower Kent...

    1995

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.