Pith. sign in

REVIEW 3 major objections 3 minor 1 cited by

Dense ReLU Neural Networks for Temporal-spatial Model

T0 review · 3 major / 3 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 fully connected ReLU networks attain minimax-optimal rates, up to logarithmic factors, for nonparametric regression under temporal dependence and function-valued spatial noise.

desk verdict Plausible first rates for dense ReLU nets with temporal-spatial dependence and spatial noise, but the main theorem is proven only under an unstated Gaussian-process assumption on gamma_i. read the letter →

arxiv 2411.09961 v8 pith:5VXLEFJN submitted 2024-11-15 stat.ML cs.LGmath.STstat.TH

Carlos Misael Madrid Padilla , Zhi Zhang , Xiaokai Luo , Daren Wang , Oscar Hernan Madrid Padilla This is my paper · ORCID
classification stat.MLcs.LGmath.STstat.TH MSC 62G0862M1062M3068T07
keywords nonparametricregressiondeepReLUnetworksbeta-mixingdependencefunctionalspatialnoisehierarchicalcompositionmodelsmanifoldlearningminimaxratestemporal-spatialdata
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 proves a convergence-rate guarantee for fully connected ReLU networks used as nonparametric regressors when the data are correlated in time and corrupted by function-valued spatial noise. The model is $y_{ij}=f^*(x_{ij})+\gamma_i(x_{ij})+\epsilon_{ij}$, with time indexed by $i$; the time series is allowed to be $\beta$-mixing while $\gamma_i$ is a spatial noise function. The main result is that a truncated least-squares network estimator achieves squared $L_2$ error of order $\max_{(p,K)\in\mathcal P}(nm)^{-2p/(2p+K)}\log^{7.3}(nm)+\log(nm)/n$ for target functions in a hierarchical composition class, and a matching lower bound shows this is minimax optimal up to logarithmic factors. This matters because it extends neural-network regression theory from independent or time-only data to spatial-temporal data, and it shows the effective complexity is the structure of the target function rather than the ambient dimension. When the inputs lie on a $d^*$-dimensional manifold, the ambient dimension is replaced by $d^*$, recovering known manifold rates when spatial noise is absent.

What carries the argument

The load-bearing device is a coupling inequality (Theorem C.17) that controls the difference between the empirical norm over the observed design and the population $L_2$ norm under $\beta$-mixing dependence. A blocking scheme cuts the $n$ time indices into alternating blocks of length $S\asymp \log n$, produces independent 'ghost' copies of each block via a standard $\beta$-mixing coupling theorem, and then applies sub-Gaussian concentration and Rademacher/covering-number bounds inside the blocks; the cost of the coupling is paid by the mixing coefficient $\beta(S)$. The approximation part of the proof rests on the ReLU-network approximation theorem for hierarchical composition models (Lemma B.1), which yields the rate-determining factor $\phi_{nm}=\max_{(p,K)\in\mathcal P}(nm)^{-2p/(2p+K)}$, and the truncation at $A_{nm}\asymp\max\{\sigma_\epsilon,\sigma_\gamma\}\log^{1/2}(nm)$ makes the response bounded on a high-probability event.

What would settle it

Construct a centered random field $\gamma$ on $[0,1]^d$ satisfying $\sup_x E\exp(c\gamma(x)^2)<\infty$ and $E\gamma(x)=0$ for all $x$, but with $\sup_x |\gamma(x)|$ tail heavier than $\exp(-Ct^2)$, for example by layering many independent sub-Gaussian components so the supremum grows with the number of components; then check whether $P(\max_{1\le i\le n}\sup_x|\gamma_i(x)|>A_{nm})$ stays bounded away from zero. If yes, Lemma B.2 fails and the uniform event in the proof of Theorem 3.1 is not guaranteed, so the rate bound as stated does not follow.

Watch

Extended reading notes

Core claim

The central claim is that dense ReLU networks can estimate $f^*$ from $n$ weakly dependent time blocks with $m$ spatial locations each at essentially the same rate as if all $nm$ points were independent, except that the spatial noise contributes an unavoidable $n^{-1}$ term. Theorem 3.1 states that with probability approaching one, $\lVert \widehat f_{A_{nm}}-f^*\rVert_{L_2}^2$ is bounded by $(\sigma_\epsilon^2+\sigma_\gamma^2+1)\max_{(p,K)\in\mathcal P}(nm)^{-2p/(2p+K)}\log^{7.3}(nm)+(\sigma_\epsilon^2+\sigma_\gamma^2)\log(nm)/n$ for $f^*\in \mathcal H(l,\mathcal P)$. Lemma 3.2 supplies the matching lower bound $C(1/n + \max_{(p,K)\in\mathcal P}(nm)^{-2p/(2p+K)})$, so up to logarithms the estimator is minimax optimal in the hierarchical composition class. In the manifold setting, Theorem 4.1 gives $(nm)^{-2p/(2p+d^*)}$ plus $\log(nm)/n$, with $d^*$ the intrinsic dimension, and the corollaries recover the time-series-only rate and the manifold rate when the spatial noise is dropped.

Load-bearing premise

The proof needs a high-probability bound saying that the spatial noise $\gamma_i$ never exceeds a growing threshold at any location, for every time point at once; the stated assumption only controls the tail of $\gamma_i(x)$ at each fixed $x$, and without an added Gaussian-process condition or a separate supremum-concentration argument that uniform control is not guaranteed.

Editorial extensions

If this is right

  • The guarantee holds for both wide networks (logarithmic depth, growing width) and deep networks (growing depth, bounded width), so the rate is not tied to one architecture.
  • Lemma 3.2 implies that the $n^{-1}$ component in the bound is unavoidable when spatial noise is present: no estimator can average away the function-valued spatial noise faster than $1/n$ over time.
  • When the inputs lie on a $d^*$-dimensional Lipschitz manifold, the rate uses $d^*$ instead of the ambient dimension $d$, so the method avoids the curse of dimensionality for low-dimensional structure.
  • For polynomial mixing decay, the paper derives a rate of $(\sigma_\epsilon^2+\sigma_\gamma^2+1)n^{-\alpha}\log^{6.3}(nm)+\sigma_\gamma^2/n$ when $m$ is large enough, showing the technique extends beyond exponential mixing.

Reading between the lines

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

  • Implicit in the proof but not stated as a theorem: the blocking-plus-coupling scheme is modular, so the same rates should transfer to other weak-dependence notions as long as the mixing coefficients decay fast enough; the polynomial-decay result is one instance.
  • A reader can test whether the $\log^{7.3}(nm)$ exponent is a proof artifact: the paper gives no matching logarithmic lower bound, so sharpening the chaining argument or proving a log lower bound would settle the question.
  • The main technical gap is not in the rate calculation: Lemma B.2 obtains the uniform spatial-noise bound with Gaussian-process tools, while Assumption 3.1(e) is only pointwise exponential-square integrability, so a complete proof as written seems to require an added Gaussian-process assumption or a separate sup-norm concentration argument.
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

3 major / 3 minor

Summary. The paper studies fully connected ReLU network estimators for the nonparametric temporal-spatial model y_ij = f^*(x_ij) + γ_i(x_ij) + ε_ij, where x_ij are design points in [0,1]^d, ε_ij are sub-Gaussian measurement errors, γ_i are functional spatial noise processes, and the temporal sequences are β-mixing with exponentially decaying coefficients. The main result (Theorem 3.1) claims that the truncated least-squares estimator achieves, with probability approaching one, an L2 error of order (σ_ε^2 + σ_γ^2 + 1) max_{(p,K)∈P} (1/(nm))^{2p/(2p+K)} log^{7.3}(nm) + (σ_ε^2 + σ_γ^2) log(nm)/n, and Lemma 3.2 claims a matching lower bound of order 1/n + max_{(p,K)∈P} (nm)^{-2p/(2p+K)}, so that the rate is minimax up to logarithmic factors. A manifold extension (Theorem 4.1) and extensive numerical comparisons are also provided.

Significance. If fully proven, the results would be a meaningful extension of the approximation-theoretic neural network rates of Kohler and Langer (2021) and Ma and Safikhani (2022) to a setting with functional spatial noise and short-range temporal dependence. The blocking construction for β-mixing sequences and the empirical-process bounds for the γ_i term are potentially reusable. The paper is also transparent about relying on external approximation theorems and does not fit parameters in the theoretical derivation; the simulations are extensive, though no code is provided. However, the main theorem is currently proven only under an unstated Gaussian-process assumption on γ_i, and the lower-bound proof contains a probability error and an insufficiently specified reduction. The central claims therefore need repair before the results can be taken as established.

major comments (3)
  1. [Appendix B, Lemma B.2] Assumption 3.1e only imposes Eγ_i(x) = 0 and sup_x E exp(c γ_i(x)^2) < ∞. The proof of the γ_i tail bound in Lemma B.2 instead assumes that γ_i is a separable centered Gaussian process on [0,1]^d and applies Borell's inequality and Fernique's theorem. This is a strictly stronger condition than Assumption 3.1e. A pointwise sub-Gaussian process can have an identically infinite sup norm: take γ_i(x) = Z_{k(x)} with independent N(0,σ^2) variables on a countable partition of [0,1]^d; then sup_x E exp(c γ_i(x)^2) < ∞ for c < 1/(2σ^2), yet sup_{x∈[0,1]^d} |γ_i(x)| = ∞ almost surely. Since the event E = {|ε_ij| ≤ A_nm/4, ||γ_i||_∞ ≤ A_nm/4} is used in Step 1 to justify the truncation inequality and in Step 4 to control the γ_i empirical-process terms, P(E) = 1 - o(1) is not available under Assumption 3.1 as stated, and Theorem 3.1 is not proven. The theorem should either be stated under an added sample-path or Gaussian assumption on γ_i, or the proof needs a genuine sup-norm concentration argument from pointwise sub-Gaussianity plus additional regularity.
  2. [Appendix G, Lemma 3.2 and Lemma G.1] The lower-bound proof claims that the ghost-coupling event Ω_x (and similarly Ω_γ, Ω_ε) 'happens with probability at least 1/n'. The union-bound argument gives P(Ω_x) ≥ 1 - n β_x(S) ≥ 1 - 1/n. The subsequent step P({||ef - f^*||^2 ≥ C/n} ∩ Ω) > C0 requires P(Ω) → 1; with only P(Ω) ≥ 1/n, the displayed inequality is not justified. This is a load-bearing error in the proof of Lemma 3.2 and Lemma G.1, and it should be corrected to 1 - o(1) or the argument restructured.
  3. [Appendix G, model (60)-(62)] The reduction to a scalar Gaussian location problem is not made rigorous. The text asserts that 'for each time i only one observation among the m_i available observations contributes' and then uses constant vectors Z_{i,l} = (z,...,z) ∈ R^{m_i} with z ∼ N(μ_l,1), but no pair of hypotheses in the hierarchical composition class H(l,P) and no spatial process γ_i satisfying Assumption 3.1 is constructed to realize this likelihood. Unless γ_i is taken to be constant in x and f^* is taken constant in the sub-problem, which is not stated, the observations y_ij = f^*(x_ij) + γ_i(x_ij) are not i.i.d. copies of a scalar N(μ_l,1) variable, and the per-block lower bound ~1/|J_e,s| does not follow from Le Cam's lemma as written. Since Lemma 3.2 is the basis for the claimed minimax optimality of the n^{-1} term, this part needs a complete construction.
minor comments (3)
  1. [Section 3.1, Eq. (9)] The displayed rate in Theorem 3.1 omits the σ_γ^2/n term that appears in the proof's final display and in the introductory display (3); because σ_γ^2 log(nm)/n dominates, this is harmless, but the two displays should be reconciled.
  2. [Appendix B, Lemma B.2] The displayed equation labeled (18) inside the lemma statement appears to be a leftover from equation (17) and is never used; it should be removed.
  3. [Proof of Theorem 3.1, Step 4] The gamma empirical-process inequality is cited as 'Theorem C.17', but the relevant statement appears to be Corollary C.10, with Lemma C.17 used for the norm coupling; the cross-references should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's rates follow from external approximation theorems and independent empirical-process arguments, with no fitted parameter renamed as a prediction.

full rationale

This is a theory paper whose upper bound (Theorem 3.1) is assembled from (i) the hierarchical-composition approximation theorem of Kohler and Langer (2021), quoted explicitly in Lemma B.1; (ii) VC-dimension/pseudodimension bounds of Bartlett et al. (2019), used in Lemma H.1; (iii) standard symmetrization, peeling, and Rademacher complexity arguments (Corollaries C.2-C.10); and (iv) a beta-mixing blocking lemma attributed to Doukhan (2012) (Theorem C.12). The lower bound (Lemma 3.2) invokes Schmidt-Hieber (2020) for the no-spatial-noise component and then runs a Le Cam argument on independent even/odd blocks for the n^{-1} term; neither ingredient is the paper's own result. No parameter is fitted to a data subset and then relabeled as a prediction: the proof chooses L, r, A_nm, and S from approximation-error and mixing conditions, not from the target L2 rate. The target rate is not encoded in Assumption 3.1 or in the least-squares objective. The only substantive caveat found is that Lemma B.2 assumes gamma_i is a separable centered Gaussian process to obtain sup-norm concentration, while Assumption 3.1(e) only states pointwise sub-Gaussian tails; this is an assumption-support gap that may invalidate the proof as written, but it is a correctness issue rather than circularity, because the missing Gaussian condition is stronger than the stated assumption and is not equivalent to the claimed rate. Self-citations (Padilla et al. 2018, 2022) appear only as background references or empirical competitors and are not load-bearing in the derivation. Accordingly, the circularity score is 0.

Assumptions & free parameters 0 free parameters · 9 assumptions · 0 invented entities

The central claim rests on standard approximation and empirical-process results plus strong dependence assumptions. The main unstated extra input is the Gaussian process assumption used in Lemma B.2 to control the sup-norm tails of gamma_i; this is the most fragile element. No new physical or statistical entities are postulated; the truncation threshold A_nm is an analytic device, not a fitted parameter.

assumptions (9)
  • domain assumption Beta-mixing temporal dependence with exponential decay beta(s) less than or similar to e^(-C_beta s), Assumption 3.1b
    Used to construct independent blocks via ghost variables in Appendix C.2.1 and to make the beta(S) terms vanish in the final rate.
  • domain assumption Independence between x, gamma, and epsilon, and within-day i.i.d. x, Assumption 3.1c
    Separates the three sources of randomness; needed for the deviation bounds and the coupling Lemma C.17.
  • domain assumption Sub-Gaussian measurement error epsilon with E(epsilon|x)=0, Assumption 3.1d
    Supplies the Hoeffding-type tail bounds used in Corollary C.1 and Theorem C.16.
  • domain assumption Spatial noise gamma has zero mean and pointwise sub-Gaussian tails sup_x E exp(c gamma^2) < infinity, Assumption 3.1e
    Intended to supply sup-norm concentration, but the proof of Lemma B.2 actually assumes gamma is a separable Gaussian process, a stronger unstated condition.
  • domain assumption f* belongs to the hierarchical composition class H(l,P) with uniformly bounded derivatives, Assumption 3.1i
    Enables the ReLU approximation rate phi_nm from Kohler and Langer (2021), stated as Lemma B.1.
  • domain assumption The design lies on a d*-dimensional Lipschitz manifold, Assumption 4.1
    Used only in Theorem 4.1 to replace ambient dimension d by intrinsic dimension d* via the approximation result of Kohler et al. (2023), stated as Lemma F.1.
  • standard math ReLU network approximation of H(l,P) with the stated L and r, Kohler and Langer (2021) Theorem 3
    Quoted as Lemma B.1; the paper relies on it for the approximation error phi_nm.
  • standard math VC-dimension and metric entropy bounds for piecewise linear networks, Bartlett et al. (2019) and Lemma H.1
    Used to bound covering numbers of the network class in the empirical process arguments.
  • standard math Minimax lower bound for hierarchical composition without spatial noise, Schmidt-Hieber (2020)
    Used as the external benchmark in Lemma G.1 to build the lower bound for the spatial-noise model.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Dense ReLU Neural Networks for Temporal-spatial Model." pith.science (2026). https://pith.science/paper/5VXLEFJN

@misc{pith2026241109961,
  author       = {Pith},
  title        = {Pith review of: Dense ReLU Neural Networks for Temporal-spatial Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5VXLEFJN}},
  note         = {Machine review of arXiv:2411.09961}
}
read the original abstract

In this paper, we focus on fully connected deep neural networks utilizing the Rectified Linear Unit (ReLU) activation function for nonparametric estimation. We derive non-asymptotic bounds that lead to convergence rates, addressing both temporal and spatial dependence in the observed measurements. By accounting for dependencies across time and space, our models better reflect the complexities of real-world data, enhancing both predictive performance and theoretical robustness. We also tackle the curse of dimensionality by modeling the data on a manifold, exploring the intrinsic dimensionality of high-dimensional data. We broaden existing theoretical findings of temporal-spatial analysis by applying them to neural networks in more general contexts and demonstrate that our proof techniques are effective for models with short-range dependence. Our empirical simulations across various synthetic response functions underscore the superior performance of our method, outperforming established approaches in the existing literature. These findings provide valuable insights into the strong capabilities of dense neural networks (Dense NN) for temporal-spatial modeling across a broad range of function classes.

Figures

Figures reproduced from arXiv: 2411.09961 by the authors.

Figure 1
Figure 1. Comparison of methods for Scenario 1 and Scenario 2 with d = 7 (first two rows), and d = 10 (last two rows), by box-plot. Each (n, mmult) setting presents the errors of the competitors within 5 times the smallest error. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_1.png] view at source ↗
Figure 2
Figure 2. Comparison of methods for Scenario 3 and Scenario 4 with d = 7 (first two rows), and d = 10 (last two rows), by box-plot. Each (n, mmult) setting presents the errors of the competitors within 5 times the smallest error [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Comparison of methods for Scenario 5 and Scenario 6 with d = 7 (first two rows), and d = 10 (last two rows), by box-plot. Each (n, mmult) setting presents the errors of the competitors within 5 times the smallest error. 15 [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Visualization of predicted ozone levels for the 45 monitoring locations on March 3rd. Subplots (a) show the y, the observed ozone levels, while subplots (b) to (f) represent predictions from Dense NN, GAM, KNN-FL, RKHS, and Trend Filtering, respectively [PITH_FULL_IMA…
Figure 5
Figure 5. Figure 5: The true mean f ∗ (xij ) and the predicted ˆf(xij ) for Scenarios 1, 2, 3 with n = 500, mmult = 1, d = 2, as estimated by Dense NN, GAM, Trend Filtering, RKHS, and KNN-FL, respectively. 83 [PITH_FULL_IMAGE:figures/full_fig_p083_5.png]
Figure 6
Figure 6. Figure 6: The true mean f ∗ (xij ) and the predicted ˆf(xij ) for Scenarios 4, 5, 6 with n = 500, mmult = 1, d = 2, as estimated by Dense NN, GAM, Trend Filtering, RKHS, and KNN-FL, respectively. 84 [PITH_FULL_IMAGE:figures/full_fig_p084_6.png]
Figure 7
Figure 7. Figure 7: Comparison of methods across different scenarios, d = 2, n, and mmult by box￾plot. For each (n, mmult) setting, we present the errors of the competitors in the range of within 5 times the smallest error. 90 [PITH_FULL_IMAGE:figures/full_fig_p090_7.png]
Figure 8
Figure 8. Figure 8: Comparison of methods across different scenarios, d = 5, n, and mmult by box￾plot. For each (n, mmult) setting, we present the errors of the competitors in the range of within 5 times the smallest error. 91 [PITH_FULL_IMAGE:figures/full_fig_p091_8.png]

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. Confidence Interval Construction and Conditional Variance Estimation with Dense ReLU Networks

    stat.ML 2024-12 conditional novelty 6.0 of 10

    Residual-based conditional variance estimation with dense ReLU networks achieves non-asymptotic rates under sub-Exponential noise, and a bootstrap interval for the conditional mean is proven to have coverage at least 1-alpha.

Reference graph

Works this paper leans on

49 extracted references · 25 canonical work pages · cited by 1 Pith paper

  1. [1]

    Nearly-tight vc-dimension and pseudodimension bounds for piecewise linear neural networks

    Peter L Bartlett, Nick Harvey, Christopher Liaw, and Abbas Mehrabian. Nearly-tight vc-dimension and pseudodimension bounds for piecewise linear neural networks. The Journal of Machine Learning Research, 20 0 (1): 0 2285--2301, 2019

    2019

  2. [2]

    On deep learning as a remedy for the curse of dimensionality in nonparametric regression

    Benedikt Bauer and Michael Kohler. On deep learning as a remedy for the curse of dimensionality in nonparametric regression . The Annals of Statistics, 47 0 (4): 0 2261 -- 2285, 2019. doi:10.1214/18-AOS1747. URL https://doi.org/10.1214/18-AOS1747

    doi:10.1214/18-aos1747 2019

  3. [3]

    Concentration Inequalities: A Nonasymptotic Theory of Independence

    Stéphane Boucheron, Gábor Lugosi, and Pascal Massart. Concentration Inequalities: A Nonasymptotic Theory of Independence . Oxford University Press, 02 2013. ISBN 9780199535255. doi:10.1093/acprof:oso/9780199535255.001.0001. URL https://doi.org/10.1093/acprof:oso/9780199535255.001.0001

    arXiv 2013

  4. [4]

    Prediction in functional linear regression

    T Tony Cai and Peter Hall. Prediction in functional linear regression. 2006

    work page 2006

  5. [5]

    Minimax and adaptive prediction for functional linear regression

    T Tony Cai and Ming Yuan. Minimax and adaptive prediction for functional linear regression. Journal of the American Statistical Association, 107 0 (499): 0 1201--1216, 2012

    work page 2012

  6. [6]

    Functional linear model

    Herv \'e Cardot, Fr \'e d \'e ric Ferraty, and Pascal Sarda. Functional linear model. Statistics & Probability Letters, 45 0 (1): 0 11--22, 1999

    work page 1999

  7. [7]

    Testing hypotheses in the functional linear model

    Herv \'e Cardot, Fr \'e d \'e ric Ferraty, Andr \'e Mas, and Pascal Sarda. Testing hypotheses in the functional linear model. Scandinavian Journal of Statistics, 30 0 (1): 0 241--255, 2003

    work page 2003

  8. [8]

    On total variation minimization and surface evolution using parametric maximum flows

    Antonin Chambolle and J \'e r \^o me Darbon. On total variation minimization and surface evolution using parametric maximum flows. International journal of computer vision, 84 0 (3): 0 288--307, 2009

    work page 2009

Show all 49 references
  1. [9]

    Nonparametric regression on low-dimensional manifolds using deep relu networks: Function approximation and statistical recovery

    Minshuo Chen, Haoming Jiang, Wenjing Liao, and Tuo Zhao. Nonparametric regression on low-dimensional manifolds using deep relu networks: Function approximation and statistical recovery. Information and Inference: A Journal of the IMA, 11 0 (4): 0 1203--1253, 2022

    2022

  2. [10]

    A deep network construction that adapts to intrinsic dimensionality beyond the domain

    Alexander Cloninger and Timo Klock. A deep network construction that adapts to intrinsic dimensionality beyond the domain. Neural Networks, 141: 0 404--419, 2021

    2021

  3. [11]

    Rates of convergence for nearest neighbor procedures

    Thomas M Cover. Rates of convergence for nearest neighbor procedures. In Proceedings of the Hawaii International Conference on Systems Sciences, volume 415, 1968

    1968

  4. [12]

    High-dimensional data analysis: The curses and blessings of dimensionality

    David L Donoho. High-dimensional data analysis: The curses and blessings of dimensionality. AMS Math Challenges Lecture, 2000

    2000

  5. [13]

    Mixing: properties and examples, volume 85

    Paul Doukhan. Mixing: properties and examples, volume 85. Springer Science & Business Media, 2012

    2012

  6. [14]

    Understanding the difficulty of training deep feedforward neural networks

    Xavier Glorot and Yoshua Bengio. Understanding the difficulty of training deep feedforward neural networks. In Proceedings of the thirteenth international conference on artificial intelligence and statistics, pages 249--256. JMLR Workshop and Conference Proceedings, 2010

    2010

  7. [15]

    Generative adversarial nets

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. Advances in neural information processing systems, 27, 2014

    2014

  8. [16]

    Long short-term memory

    Alex Graves and Alex Graves. Long short-term memory. Supervised sequence labelling with recurrent neural networks, pages 37--45, 2012

    2012

  9. [17]

    o rfi, Michael K \

    L \'a szl \'o Gy \"o rfi, Michael K \"o hler, Adam Krzy \.z ak, and Harro Walk. A distribution-free theory of nonparametric regression, volume 1. Springer, 2002

    2002

  10. [18]

    Methodology and convergence rates for functional linear regression

    Peter Hall and Joel L Horowitz. Methodology and convergence rates for functional linear regression. 2007

    2007

  11. [19]

    Nonasymptotic bounds on the l 2 error of neural network regression estimates

    Michael Hamers and Michael Kohler. Nonasymptotic bounds on the l 2 error of neural network regression estimates. Annals of the Institute of Statistical Mathematics, 58: 0 131--151, 2006

    2006

  12. [20]

    Kernel methods in machine learning

    Thomas Hofmann, Bernhard Sch \"o lkopf, and Alexander J Smola. Kernel methods in machine learning. 2008

    2008

  13. [21]

    Deep nonparametric regression on approximate manifolds: Nonasymptotic error bounds with polynomial prefactors

    Yuling Jiao, Guohao Shen, Yuanyuan Lin, and Jian Huang. Deep nonparametric regression on approximate manifolds: Nonasymptotic error bounds with polynomial prefactors. The Annals of Statistics, 51 0 (2): 0 691--716, 2023

    2023

  14. [22]

    Adaptive regression estimation with multilayer feedforward neural networks

    Michael Kohler and Adam Krzy \.z ak. Adaptive regression estimation with multilayer feedforward neural networks. Nonparametric Statistics, 17 0 (8): 0 891--913, 2005

    2005

  15. [23]

    On the rate of convergence of fully connected very deep neural network regression estimates

    Michael Kohler and Sophie Langer. On the rate of convergence of fully connected very deep neural network regression estimates. arXiv preprint arXiv:1908.11133, 2019

    1908 arXiv

  16. [24]

    On the rate of convergence of fully connected deep neural network regression estimates

    Michael Kohler and Sophie Langer. On the rate of convergence of fully connected deep neural network regression estimates. The Annals of Statistics, 49 0 (4): 0 2231--2249, 2021

    2021

  17. [25]

    Estimation of a regression function on a manifold by fully connected deep neural networks

    Michael Kohler, Sophie Langer, and Ulrich Reif. Estimation of a regression function on a manifold by fully connected deep neural networks. Journal of Statistical Planning and Inference, 222: 0 160--181, 2023

    2023

  18. [26]

    Imagenet classification with deep convolutional neural networks

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks. Advances in neural information processing systems, 25, 2012

    2012

  19. [27]

    Imagenet classification with deep convolutional neural networks

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks. Communications of the ACM, 60 0 (6): 0 84--90, 2017

    2017

  20. [28]

    Maximal inequalities and some applications

    Franziska K \"u hn and Ren \'e L Schilling. Maximal inequalities and some applications. Probability Surveys, 20: 0 382--485, 2023

    2023

  21. [29]

    Statistical methods in spatial epidemiology

    Andrew B Lawson. Statistical methods in spatial epidemiology. John Wiley & Sons, 2013

    2013

  22. [30]

    Theoretical analysis of deep neural networks for temporally dependent observations

    Mingliang Ma and Abolfazl Safikhani. Theoretical analysis of deep neural networks for temporally dependent observations. Advances in Neural Information Processing Systems, 35: 0 37324--37334, 2022

    2022

  23. [31]

    Convergence rates for single hidden layer feedforward networks

    Daniel F McCaffrey and A Ronald Gallant. Convergence rates for single hidden layer feedforward networks. Neural Networks, 7 0 (1): 0 147--158, 1994

    1994

  24. [32]

    Rectified linear units improve restricted boltzmann machines

    Vinod Nair and Geoffrey E Hinton. Rectified linear units improve restricted boltzmann machines. In Proceedings of the 27th international conference on machine learning (ICML-10), pages 807--814, 2010

    2010

  25. [33]

    Adaptive non-parametric regression with the k -nn fused lasso

    Oscar Hernan Madrid Padilla, James Sharpnack, Yanzhen Chen, and Daniela M Witten. Adaptive non-parametric regression with the k -nn fused lasso. arXiv preprint arXiv:1807.11641, 2018

    2018 arXiv

  26. [34]

    Quantile regression with relu networks: Estimators and minimax rates

    Oscar Hernan Madrid Padilla, Wesley Tansey, and Yanzhen Chen. Quantile regression with relu networks: Estimators and minimax rates. Journal of Machine Learning Research, 23 0 (247): 0 1--42, 2022

    2022

  27. [35]

    Functional data analysis for density functions by transformation to a hilbert space

    Alexander Petersen and Hans-Georg M \"u ller. Functional data analysis for density functions by transformation to a hilbert space. 2016

    2016

  28. [36]

    Improving language understanding by generative pre-training

    A Radford. Improving language understanding by generative pre-training. 2018

    2018

  29. [37]

    Fast and flexible admm algorithms for trend filtering

    Aaditya Ramdas and Ryan J Tibshirani. Fast and flexible admm algorithms for trend filtering. Journal of Computational and Graphical Statistics, 25 0 (3): 0 839--858, 2016

    2016

  30. [38]

    Nonlinear dimensionality reduction by locally linear embedding

    ST Roweis and LK Saul. Nonlinear dimensionality reduction by locally linear embedding. Science, 290 0 (5500): 0 2323--2326, 2000

    2000

  31. [39]

    Additive models with trend filtering

    Veeranjaneyulu Sadhanala and Ryan J Tibshirani. Additive models with trend filtering. The Annals of Statistics, 47 0 (6): 0 3032--3068, 2019

    2019

  32. [40]

    Nonparametric regression using deep neural networks with relu activation function

    Johannes Schmidt-Hieber. Nonparametric regression using deep neural networks with relu activation function. 2020

    2020

  33. [41]

    Optimal global rates of convergence for nonparametric regression

    Charles J Stone. Optimal global rates of convergence for nonparametric regression. The annals of statistics, pages 1040--1053, 1982

    1982

  34. [42]

    Additive regression and other nonparametric models

    Charles J Stone. Additive regression and other nonparametric models. The annals of Statistics, 13 0 (2): 0 689--705, 1985

    1985

  35. [43]

    The use of polynomial splines and their tensor products in multivariate function estimation

    Charles J Stone. The use of polynomial splines and their tensor products in multivariate function estimation. The annals of statistics, 22 0 (1): 0 118--171, 1994

    1994

  36. [44]

    Deepface: Closing the gap to human-level performance in face verification

    Yaniv Taigman, Ming Yang, Marc'Aurelio Ranzato, and Lior Wolf. Deepface: Closing the gap to human-level performance in face verification. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1701--1708, 2014

    2014

  37. [45]

    A global geometric framework for nonlinear dimensionality reduction

    JB Tenenbaum, V De Silva, and JC Langford. A global geometric framework for nonlinear dimensionality reduction. Science, 290 0 (5500): 0 2319--2323, 2000

    2000

  38. [46]

    Attention is all you need

    A Vaswani. Attention is all you need. Advances in Neural Information Processing Systems, 2017

    2017

  39. [47]

    Spatio-temporal statistics with R

    Christopher K Wikle, Andrew Zammit-Mangion, and Noel Cressie. Spatio-temporal statistics with R. CRC Press, 2019

    2019

  40. [48]

    Generalized additive models: an introduction with R

    Simon N Wood. Generalized additive models: an introduction with R. chapman and hall/CRC, 2017

    2017

  41. [49]

    The varying driving forces of urban land expansion in china: Insights from a spatial-temporal analysis

    Rong Wu, Zhigang Li, and Shaojian Wang. The varying driving forces of urban land expansion in china: Insights from a spatial-temporal analysis. Science of The Total Environment, 766: 0 142591, 2021

    2021

Pith tools

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