Pith. sign in

REVIEW 4 major objections 6 minor 54 references

Explainable Neural Networks with Guarantees: A Sparse Estimation Approach

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read SparXnet proves that the sample complexity of an explainable neural network depends linearly on the number of selected features and only logarithmically on the input dimension, with no dependence on network width, depth, or parameter…

desk verdict The generalization bound is new and likely correct for an idealized Lipschitz class, but the paper never shows the trained SparXnet lives in that class, so the headline 'guarantee' is not actually delivered. read the letter →

arxiv 2501.02010 v2 pith:E6RXDC64 submitted 2025-01-02 cs.LG

classification cs.LG MSC 68T0768Q32
keywords SparXnetexplainableneuralnetworkssparsefeatureselectionsoftmaxroutingsamplecomplexitygeneralizationboundsLipschitzfunctionsadditivemodels
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper introduces SparXnet, a neural network whose prediction is a linear combination of a small number of learned one-dimensional functions, each applied to a single input feature chosen by a softmax-gated first layer. The central claim is a generalization bound: to reach excess risk $\epsilon$ the number of samples needed is $O(K L^2 / \epsilon^2 \log^3(K L^2 \log(d+L+1)/\epsilon^2))$, linear in the number of selected features $K$, logarithmic in the total number of features $d$, and independent of the width, depth, or parameter count of the networks that implement the one-dimensional functions, provided those functions are $L$-Lipschitz. If true, this means a user can add an arbitrary number of irrelevant features or enlarge the per-feature networks without paying a sample-complexity penalty, while the final model remains inspectable by plotting the learned $f_k$. The paper supports the guarantee experimentally on synthetic data and six real-world datasets, showing competitive accuracy and successful recovery of the true sparse features.

What carries the argument

SparXnet's key object is the softmax-gated selection layer: each pathway $k$ has a weight vector $w^k$ whose softmax $W^k = \exp(w^k/\tau) / \sum_v \exp(w^k_v/\tau)$ concentrates on one feature as temperature $\tau$ decreases, turning the first layer into a soft routing mechanism that selects a single input dimension per pathway. The selected one-dimensional input is then passed through a trainable $L$-Lipschitz function $f_k$, implemented as a deep fully connected network, and the $K$ outputs are combined in a final linear layer. What carries the theoretical argument is a covering-number bound for the resulting function class: the $d$-dimensional selection vectors are covered by a net of size roughly $(12\chi L d N / \epsilon)^{O(K L^2 \chi^2 \Gamma^2 / \epsilon^2)}$ via Zhang's bound for $\ell^1$-bounded linear classes, and each $f_k$ is covered by a net of $(4\chi L / \epsilon)^{O(\chi L \Gamma / \epsilon)}$ using the low-dimensional Lipschitz covering bound of von Luxburg and Bousquet; Dudley's entropy integral converts these log-covering numbers into an empirical Rademacher complexity bound of order $(1/\sqrt{N})(\chi L \Gamma \sqrt{K} + 1) \sqrt{\log(dN^2)} \log N$. Because the per-feature functions have one-dimensional inputs, the covering cost depends only on the Lipschitz constant $L$ and the interval length, never on the number of parameters of the networks representing $f_k$.

What would settle it

Take a synthetic sparse regression with known ground truth and a fixed small $K$, train SparXnet with a prescribed $L$, then numerically estimate the Lipschitz constant of each learned $f_k$ on a dense grid and measure test excess risk as $N$ grows; if the realized Lipschitz constants substantially exceed $L$, or if the excess risk does not track $O(\sqrt{K L^2 \log(d)/N})$, then the claim that the theorem governs the trained model fails.

Watch

Extended reading notes

Core claim

SparXnet's function class is $F(x) = \beta + \sum_{k=1}^K \theta_k f_k(\sum_{u=1}^d W^k_u x_u)$, where $W^k = \mathrm{softmax}(w^k/\tau)$ is a learned probability distribution over the $d$ input features, $f_k: \mathbb{R} \to \mathbb{R}$ is an $L$-Lipschitz function represented by a deep network, and the final weights satisfy $\sum_k |\theta_k| \leq \Gamma$. The paper proves that for a bounded Lipschitz loss, empirical-risk minimization over this class has excess risk bounded by $O((L/\sqrt{N})(\chi L \Gamma \sqrt{K} + 1) \sqrt{\log(12 d N^2 [\chi L \Gamma + 1])} \log N + B\sqrt{\log(1/\delta)/N})$, and hence that the sample size required for excess risk $\epsilon$ is $O(K L^2 / \epsilon^2 \log^3(K L^2 \log(d+L+1)/\epsilon^2))$ when the constants are $O(1)$. The result is built on covering-number estimates: the softmax rows live in the simplex of dimension $d$, whose covering cost is $O(\log d)$, and each one-dimensional $L$-Lipschitz function contributes a covering cost depending only on $L$, not on the network architecture. The authors take this to show that sparse feature selection from high-dimensional inputs is almost free in sample complexity, and that interpretable one-dimensional feature effects can be learned with guarantees.

Load-bearing premise

The guarantee assumes the one-dimensional functions $f_k$ are $L$-Lipschitz with a fixed known constant $L$, but training never enforces or verifies this, so the bound may not apply to the actual trained networks.

Editorial extensions

If this is right

  • Adding irrelevant features to the input costs only a logarithmic factor in the sample bound, so SparXnet can be applied directly to high-dimensional tabular data with few true signals.
  • The size and depth of the networks that implement each one-dimensional $f_k$ can be increased freely without worsening the generalization guarantee, as long as each network stays $L$-Lipschitz.
  • The same sample-complexity statement covers both regression (truncated square loss) and binary classification (logistic loss), so the interpretability guarantee transfers to common classification benchmarks.
  • Because the final prediction is a linear combination of one-dimensional functions, a user can plot each $f_k$ to read off thresholds and nonlinear effects, such as a sharp jump in default risk above a 20 percent interest rate.
  • The bound's independence from the number of parameters distinguishes SparXnet from standard neural-network generalization bounds, which typically scale with parameter count or weight norms.

Reading between the lines

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

  • The training procedure does not constrain or certify the Lipschitz constant $L$ of the networks $f_k$, and the paper never measures the realized Lipschitz constant; enforcing a Lipschitz bound during training (e.g., spectral normalization) would make the proven bound directly applicable to the deployed model.
  • A testable extension suggested by the bound: on a synthetic problem with a fixed small $K$ and increasing $d$, the excess risk should grow roughly like $\sqrt{\log d}/\sqrt{N}$; measuring this scaling would separate the effect of feature selection from the effect of input dimensionality.
  • The temperature annealing schedule is heuristic: no proof is given that the softmax weights saturate to a hard one-hot selection during training, so the interpretability guarantee in practice depends on weight saturation holding; proving or regularizing for saturation would close the gap between architecture description and the linear-combination interpretation.
  • The proof technique, which charges only $O(\log d)$ for feature selection and $O(1)$ per one-dimensional Lipschitz function, could transfer to other structured parsimonious models such as sparse additive or knot-based models, where each component is low-dimensional and smooth.
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

4 major / 6 minor

Summary. The paper introduces SparXnet, an additive neural network architecture that learns K one-dimensional feature transformations applied to softmax-weighted selections of input features, with the final prediction being a linear combination of these transformations. The authors prove a generalization bound (Theorem 1) and a sample-complexity corollary (Corollary 2) that they claim is linear in the number of selected features K and logarithmic in the input dimension d, with no dependence on the number of network parameters under a Lipschitz continuity assumption. They report experiments on synthetic and real-world datasets showing competitive predictive performance and interpretable feature effects. The central theoretical claims, however, rely on an L-Lipschitz assumption that is neither enforced nor verified for the trained networks, and the proof of the covering bound contains a gap.

Significance. If the theoretical results were correct and applicable to the trained model, SparXnet would be a valuable contribution to interpretable machine learning: the architecture is simple and directly interpretable, the empirical evaluation across six real datasets is useful, and a parameter-count-free generalization bound for an additive feature-selection network would be of genuine interest. The paper also correctly identifies that 1D Lipschitz function classes have low metric entropy and that this can be exploited. However, as written, the gap between the proven bound and the trained model, together with the technical flaws in the covering proof and the L-exponent inconsistency, prevents the advertised guarantee from being substantiated. The empirical work alone is solid but does not by itself support the title's 'guarantees'.

major comments (4)
  1. [Section 4 / Training procedure] The generalization guarantee in Theorem 1 applies to the ERM over the class F in Proposition 7, which requires every f_k to be L-Lipschitz for a known constant L. SparXnet as implemented optimizes unconstrained fully connected networks for f_k and never enforces, estimates, or reports Lipschitz constants (Section 5 reports no such measurement, and the appendix's limitation paragraph mentions gradient regularization only as a possible future improvement). Consequently, the trained network need not belong to F, and the proven bound does not apply to the model actually deployed. This is load-bearing for the paper's central 'guarantees' claim.
  2. [Appendix A.1, Eq. (13)] The proof of Proposition 7 has a gap in the covering argument. Equation (13) asserts |⟨x, W^k − W̄^k⟩| ≤ (ϵ/(3Lχ)) χ Σ_u W^k_u, but the coordinatewise cover condition (11) only gives a bound on each |W^k_u − W̄^k_u|; the step replacing Σ_u |W^k_u − W̄^k_u| by Σ_u W^k_u is unjustified. Without a uniform bound on the inner product over the sample, the cover C constructed in the proof is not guaranteed to be an ϵ-cover of F, so the stated log d dependence in Eq. (10) is not established by the provided proof.
  3. [Section 4, Theorem 1 vs Corollary 2] There is an inconsistency in the dependence on L between Theorem 1 and Corollary 2. The leading term in Eq. (5) is L · (L √K)/√N, so after setting excess risk to ε the sample complexity scales as K L^4/ε^2 (up to log factors) when both the loss and the f_k have Lipschitz constant bounded by L. Corollary 2 and the proof in Appendix A.1 instead state K L^2/ε^2 and drop one factor of L in the excess-risk expression. Either the theorem's L outside the bracket is intended to be a constant (not the same L as the f_k) or the corollary's exponent is wrong; the current text cannot be read consistently.
  4. [Section 4, Notation] The symbol L is overloaded: it denotes both the Lipschitz constant of the loss ℓ (Theorem 1 statement) and the Lipschitz bound of the functions f_k (Proposition 7 and the introduction). This ambiguity directly affects the interpretation of Eq. (5) and the O(K L^2) claim in Corollary 2 and the abstract, and it contributes to the exponent error in the previous comment. Distinct symbols (e.g., L_loss and L_f) should be introduced and used consistently.
minor comments (6)
  1. [Appendix A.1, Eq. (13)] In the displayed expression, 'min_k' should be 'max_k', since the goal is to control the largest deviation over the K pathways.
  2. [Section 4, after Theorem 1] The phrase 'low-dimensional input (1 or 2 d)' contains a typo; it should be '1 or 2 dimensions'.
  3. [Corollary 2 and abstract] The sample-complexity statement in the abstract and introduction omits the dependence on the target excess risk ε; please state the dependence on ε consistently with Corollary 2.
  4. [Section 4] The function class F used in Theorem 1 is formally defined only in the appendix (Proposition 7); it should be stated in Section 4 before the theorem so that the main text is self-contained.
  5. [Appendix B.3] Figure 6 is discussed in the text but the heatmap is not legible in the current version; please provide a higher-resolution figure and a caption explaining which weights are saturated.
  6. [References] The reference 'Bartlet and Shawe-taylor' is incomplete; the title and venue are missing.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the SparXnet generalization analysis is self-contained given standard external metric-entropy and Rademacher bounds; self-citations are not load-bearing.

full rationale

Walking the derivation chain in Section 4 and Appendix A.1, the generalization bound is obtained by bounding covering numbers of the class F in Proposition 7, applying Dudley's entropy formula (Corollary 8), Talagrand's contraction lemma, and a standard Rademacher uniform-convergence theorem. Each external result is cited with independent sources: Lipschitz covering numbers from von Luxburg and Bousquet (2004) and Tikhomirov (1993); metric-entropy covering bounds ultimately from Zhang (2002) and Pisier/Bartlett-Shawe-Taylor; Rademacher complexity and Talagrand's lemma from Scott, Ledoux-Talagrand, and Meir-Zhang. The self-citations to Ledent et al. (2021), Vandermeulen and Ledent (2021), and Ledent and Alves (2024) are used only as pointers to standard covering-number/Rademacher inequalities or as examples of prior use of low-dimensional Lipschitz complexity, and they do not assume the target sample-complexity result. Corollary 2 is a direct algebraic rearrangement of Theorem 1's excess-risk bound, not a fitted quantity or an empirical 'prediction'. The paper's own limitation passage about softmax non-saturation and gradient regularization concerns whether the assumption that f_k are L-Lipschitz is enforced for the trained network; that is an assumption-verification gap, not circular reasoning. No step in the derivation is equivalent to its own input by construction.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The theoretical guarantee rests on the assumption that each f_k is L-Lipschitz and on standard metric entropy bounds. The empirical method introduces no new physical entities. The key unmeasured quantity is L, which is assumed but not enforced or estimated.

free parameters (3)
  • K (number of pathways)
    User-specified number of features to select; controlled in experiments to trade off sparsity and accuracy.
  • Temperature tau
    Softmax temperature tuned by Bayesian optimization and annealed during training; controls sharpness of feature selection and interpretability.
  • Dropout rate, learning rate, batch size, hidden layer sizes
    Standard hyperparameters tuned per dataset by Bayesian optimization; they affect empirical results but not the theoretical bound.
assumptions (5)
  • domain assumption The learned functions f_k are L-Lipschitz continuous for a known constant L.
    Invoked before Theorem 1; the generalization bound is for the class F of L-Lipschitz functions, but training does not enforce or verify this property.
  • domain assumption The loss function ℓ is bounded by B and has Lipschitz constant at most L.
    Stated in Theorem 1; required for Talagrand contraction and the final excess risk bound.
  • domain assumption Input features satisfy ||x||_max ≤ χ.
    Section 3 normalization assumption, used throughout the covering number argument.
  • domain assumption Softmax weights W^k are positive and sum to at most 1.
    Definition of the function class F in Proposition 7; the model uses softmax so the sum is exactly 1, and the bound covers the slightly larger class.
  • standard math Standard metric entropy bounds for Lipschitz functions and linear function classes hold (von Luxburg and Bousquet, Zhang, Vershynin).
    Used in Appendix A.1 to bound covering numbers of the 1D Lipschitz functions and of the softmax weight simplex.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Explainable Neural Networks with Guarantees: A Sparse Estimation Approach." pith.science (2026). https://pith.science/paper/E6RXDC64

@misc{pith2026250102010,
  author       = {Pith},
  title        = {Pith review of: Explainable Neural Networks with Guarantees: A Sparse Estimation Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E6RXDC64}},
  note         = {Machine review of arXiv:2501.02010}
}
read the original abstract

Balancing predictive power and interpretability has long been a challenging research area, particularly in powerful yet complex models like neural networks, where nonlinearity obstructs direct interpretation. This paper introduces a novel approach to constructing an explainable neural network that harmonizes predictiveness and explainability. Our model, termed SparXnet, is designed as a linear combination of a sparse set of jointly learned features, each derived from a different trainable function applied to a single 1-dimensional input feature. Leveraging the ability to learn arbitrarily complex relationships, our neural network architecture enables automatic selection of a sparse set of important features, with the final prediction being a linear combination of rescaled versions of these features. We demonstrate the ability to select significant features while maintaining comparable predictive performance and direct interpretability through extensive experiments on synthetic and real-world datasets. We also provide theoretical analysis on the generalization bounds of our framework, which is favorably linear in the number of selected features and only logarithmic in the number of input features. We further lift any dependence of sample complexity on the number of parameters or the architectural details under very mild conditions. Our research paves the way for further research on sparse and explainable neural networks with guarantee.

Figures

Figures reproduced from arXiv: 2501.02010 by the authors.

Figure 1
Figure 1. Schematic overview of the proposed model in the case of two selected features ( [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Visualizing the learned model predictions using 2000 noisy observations and three features, including one [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. The dynamics between predictive accuracy and recovery rate of true features as the sparsity level varies [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Illustrating the model inference for two credit applicants in the credit risk dataset. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Comparison of mean test AUC by the number of pathways across datasets. The analysis highlights the [PITH_FULL_IMAGE:figures/full_fig_p019_5.png]
Figure 6
Figure 6. Figure 6: Heatmap of weights between the input layer and the first hidden layer after the softmax operation. The figure [PITH_FULL_IMAGE:figures/full_fig_p020_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 43 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Agarwal, L

    R. Agarwal, L. Melnick, N. Frosst, X. Zhang, B. Lengerich, R. Caruana, and G. E. Hinton. Neural additive models: Interpretable machine learning with neural nets. In M. Ranzato, A. Beygelzimer, Y. Dauphin, P. Liang, and J. W. Vaughan, editors, Advances in Neural Information Processing Systems, volume 34, pages 4699--4711. Curran Associates, Inc., 2021. URL...

  4. [4]

    Arora, S

    S. Arora, S. Du, W. Hu, Z. Li, and R. Wang. Fine-grained analysis of optimization and generalization for overparameterized two-layer neural networks. In ICML, 2019

  5. [5]

    Bartlett and J

    P. Bartlett and J. Shawe-taylor. Generalization performance of support vector machines and other pattern classifiers, 1998

  6. [6]

    P. L. Bartlett, D. J. Foster, and M. J. Telgarsky. Spectrally-normalized margin bounds for neural networks. In I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, editors, Advances in Neural Information Processing Systems 30, pages 6240--6249. Curran Associates, Inc., 2017

  7. [7]

    K. W. Church. Word2vec. Natural Language Engineering, 23 0 (1): 0 155--162, 2017

  8. [8]

    Dalvi, N

    F. Dalvi, N. Durrani, H. Sajjad, Y. Belinkov, A. Bau, and J. Glass. What is one grain of sand in the desert? analyzing individual neurons in deep nlp models. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 33, pages 6309--6317, 2019

Show all 54 references
  1. [9]

    Devlin, M.-W

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018

  2. [10]

    Dhurandhar, P.-Y

    A. Dhurandhar, P.-Y. Chen, R. Luss, C.-C. Tu, P. Ting, K. Shanmugam, and P. Das. Explanations based on the missing: Towards contrastive explanations with pertinent negatives. Advances in neural information processing systems, 31, 2018

  3. [11]

    S. S. Du, X. Zhai, B. Poczos, and A. Singh. Gradient descent provably optimizes over-parameterized neural networks. In International Conference on Learning Representations, 2019

  4. [12]

    J. H. Friedman and W. Stuetzle. Projection pursuit regression. Journal of the American Statistical Association, 76 0 (376): 0 817--823, 1981

  5. [13]

    Goyal, Z

    Y. Goyal, Z. Wu, J. Ernst, D. Batra, D. Parikh, and S. Lee. Counterfactual visual explanations. In International Conference on Machine Learning, pages 2376--2384. PMLR, 2019

  6. [14]

    F. Graf, S. Zeng, B. Rieck, M. Niethammer, and R. Kwitt. On measuring excess capacity in neural networks. Advances in Neural Information Processing Systems, 35: 0 10164--10178, 2022

  7. [15]

    Hastie and R

    T. Hastie and R. Tibshirani. Generalized additive models. Statistical Science, 1 0 (3): 0 297--310, 1986

  8. [16]

    F. He, T. Liu, and D. Tao. Why resnet works? residuals generalize. IEEE transactions on neural networks and learning systems, 31 0 (12): 0 5349--5362, 2020

  9. [17]

    K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 770--778, 2016

  10. [18]

    A. Jacot. Implicit bias of large depth networks: a notion of rank for nonlinear functions. In The Eleventh International Conference on Learning Representations, 2023. URL https://openreview.net/forum?id=6iDHce-0B-a

  11. [19]

    Jacot, F

    A. Jacot, F. Gabriel, and C. Hongler. Neural tangent kernel: Convergence and generalization in neural networks. In S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett, editors, Advances in Neural Information Processing Systems, volume 31. Curran A...

  12. [20]

    Lecun, L

    Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner. Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86 0 (11): 0 2278--2324, 1998. doi:10.1109/5.726791

  13. [21]

    Ledent and R

    A. Ledent and R. Alves. Generalization analysis of deep non-linear matrix completion. In R. Salakhutdinov, Z. Kolter, K. Heller, A. Weller, N. Oliver, J. Scarlett, and F. Berkenkamp, editors, Proceedings of the 41st International Conference on Machine Learning, volume 235 of P...

  14. [22]

    Ledent, W

    A. Ledent, W. Mustafa, Y. Lei, and M. Kloft. Norm-based generalisation bounds for deep multi-class convolutional neural networks. Proceedings of the AAAI Conference on Artificial Intelligence, 35 0 (9): 0 8279--8287, May 2021. URL https://ojs.aaai.org/index.php/AAAI/article/view/17007

  15. [23]

    Ledoux and M

    M. Ledoux and M. Talagrand. Probability in Banach spaces : isoperimetry and processes. Springer, Berlin [u.a.], 1991. ISBN 3540520139. URL http://digitale-objekte.hbz-nrw.de/storage/2008/01/16/file_132/2293955.pdf

  16. [24]

    Y. Lei, U. Dogan, D.-X. Zhou, and M. Kloft. Data-dependent generalization bounds for multi-class classification. IEEE Transactions on Information Theory, 65 0 (5): 0 2995--3021, 2019. doi:10.1109/TIT.2019.2893916

  17. [25]

    P. Liu. An integrated framework on human-in-the-loop risk analytics. Journal of Financial Data Science, 5 0 (1): 0 58--64, 2023. doi:10.3905/jfds.2022.1.116. URL https://doi.org/10.3905/jfds.2022.1.116

  18. [26]

    Z. Liu, Y. Wang, S. Vaidya, F. Ruehle, J. Halverson, M. Solja c i \'c , T. Y. Hou, and M. Tegmark. Kan: Kolmogorov-arnold networks. arXiv preprint arXiv:2404.19756, 2024

  19. [27]

    P. M. Long and H. Sedghi. Size-free generalization bounds for convolutional neural networks. In International Conference on Learning Representations, 2020

  20. [28]

    Meir and T

    R. Meir and T. Zhang. Generalization error bounds for bayesian mixture algorithms. J. Mach. Learn. Res., 4 0 (null): 0 839–860, dec 2003. ISSN 1532-4435

  21. [29]

    Mustafa, Y

    W. Mustafa, Y. Lei, A. Ledent, and M. Kloft. Fine-grained generalization analysis of structured output prediction. In Z.-H. Zhou, editor, Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI-21 , pages 2841--2847. International Joint Co...

  22. [30]

    Nagarajan and J

    V. Nagarajan and J. Z. Kolter. Deterministic pac-bayesian generalization bounds for deep networks via generalizing noise-resilience. CoRR, abs/1905.13344, 2019

  23. [31]

    Maurey-Schwartz

    G. Pisier. Remarques sur un r\'esultat non publi\'e de b. maurey. S\'eminaire Analyse fonctionnelle (dit "Maurey-Schwartz"), 1980-1981. talk:5

  24. [32]

    V. Roth. The generalized lasso. IEEE transactions on neural networks, 15 0 (1): 0 16--28, 2004

  25. [33]

    C. Rudin. Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nature machine intelligence, 1 0 (5): 0 206--215, 2019

  26. [34]

    Sardy and P

    S. Sardy and P. Tseng. Amlet, ramlet, and gamlet: Automatic nonlinear fitting of additive models, robust and generalized, with wavelets. Journal of Computational and Graphical Statistics, 13 0 (2): 0 283--309, 2004

  27. [35]

    C. Scott. Rademacher complexity. Lecture Notes, Statistical Learning Theory, 2014

  28. [36]

    Simonyan, A

    K. Simonyan, A. Vedaldi, and A. Zisserman. Deep inside convolutional networks: Visualising image classification models and saliency maps. CoRR, abs/1312.6034, 2013

  29. [37]

    G. D. Stormo, T. D. Schneider, L. Gold, and A. Ehrenfeucht. Use of the ‘perceptron’algorithm to distinguish translational initiation sites in e. coli. Nucleic acids research, 10 0 (9): 0 2997--3011, 1982

  30. [38]

    Szegedy, W

    C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V. Vanhoucke, and A. Rabinovich. Going deeper with convolutions. In 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 1--9, 2015. doi:10.1109/CVPR.2015.7298594

  31. [39]

    M. TAN, Y. DAI, D. TANG, Z. FENG, G. HUANG, J. JIANG, J. LI, and S. SHI. Exploring and adapting chinese gpt to pinyin input method. Association for Computational Linguistics, 2022

  32. [40]

    R. J. Tibshirani. The lasso problem and uniqueness. 2013

  33. [41]

    V. M. Tikhomirov. -Entropy and -Capacity of Sets In Functional Spaces, pages 86--170. Springer Netherlands, Dordrecht, 1993. ISBN 978-94-017-2973-4. doi:10.1007/978-94-017-2973-4_7. URL https://doi.org/10.1007/978-94-017-2973-4_7

  34. [42]

    R. A. Vandermeulen and A. Ledent. Beyond smoothness: Incorporating low-rank analysis into nonparametric density estimation. Advances in Neural Information Processing Systems, 34: 0 12180--12193, 2021

  35. [43]

    Varshneya, A

    S. Varshneya, A. Ledent, R. A. Vandermeulen, Y. Lei, M. Enders, D. Borth, and M. Kloft. Learning interpretable concept groups in cnns. In IJCAI International Joint Conference on Artificial Intelligence, 2021

  36. [44]

    Vershynin

    R. Vershynin. High-dimensional probability. 2019. URL https://www.math.uci.edu/ rvershyn/papers/HDP-book/HDP-book.pdf

  37. [45]

    von Luxburg and O

    U. von Luxburg and O. Bousquet. Distance-based classification with lipschitz functions. J. Mach. Learn. Res., 5 0 (Jun): 0 669--695, 2004

  38. [46]

    Wachter, B

    S. Wachter, B. Mittelstadt, and C. Russell. Counterfactual explanations without opening the black box: Automated decisions and the gdpr. Harv. JL & Tech., 31: 0 841, 2017

  39. [47]

    Wei and T

    C. Wei and T. Ma. Data-dependent sample complexity of deep neural networks via lipschitz augmentation. In H. Wallach, H. Larochelle, A. Beygelzimer, F. d Alch\' e -Buc, E. Fox, and R. Garnett, editors, Advances in Neural Information Processing Systems 32, pages 9725--9736. Cur...

  40. [48]

    L. Wu, A. Ledent, Y. Lei, and M. Kloft. Fine-grained generalization analysis of vector-valued learning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 10338--10346, 2021

  41. [49]

    M. D. Zeiler and R. Fergus. Visualizing and understanding convolutional networks. CoRR, abs/1311.2901, 2013. URL http://arxiv.org/abs/1311.2901

  42. [50]

    T. Zhang. Covering number bounds of certain regularized linear function classes. Journal of Machine Learning Research, 2: 0 527--550, Mar. 2002. ISSN 1532-4435. doi:10.1162/153244302760200713

  43. [51]

    Zhang, P

    Y. Zhang, P. Ti n o, A. Leonardis, and K. Tang. A survey on neural network interpretability. IEEE Transactions on Emerging Topics in Computational Intelligence, 5 0 (5): 0 726--742, 2021

  44. [52]

    B. Zhou, A. Khosla, \` A . Lapedriza, A. Oliva, and A. Torralba. Object detectors emerge in deep scene cnns. CoRR, abs/1412.6856, 2014. URL http://arxiv.org/abs/1412.6856

  45. [53]

    B. Zhou, D. Bau, A. Oliva, and A. Torralba. Interpreting deep visual representations via network dissection. IEEE transactions on pattern analysis and machine intelligence, 2018

  46. [54]

    H. Zou. The adaptive lasso and its oracle properties. Journal of the American statistical association, 101 0 (476): 0 1418--1429, 2006

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.