Pith. sign in

REVIEW 4 major objections 5 minor 25 references

Degree-Optimized Cumulative Polynomial Kolmogorov-Arnold Networks

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that CP-KAN reduces polynomial degree selection in Chebyshev KANs to a per-layer QUBO optimization, replacing exponential search, and that this yields competitive regression performance with far fewer parameters.

desk verdict The paper's central thesis is not supported: Equation (7) is separable and would always pick the maximum degree, while the experiments must be using an unstated complexity penalty; still, the empirical work and the underlying question deserve a rigorous referee. read the letter →

arxiv 2505.15228 v1 pith:CASYMG75 submitted 2025-05-21 cs.LG cs.CEcs.NE

Mathew Vanherreweghe , Lirandë Pira , Patrick Rebentrost This is my paper
classification cs.LGcs.CEcs.NE
keywords Kolmogorov-ArnoldnetworksChebyshevpolynomialsQUBOdegreeselectionsimulatedannealingregressionfinancialtimeseriesparameterefficiency
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 cumulative polynomial Kolmogorov-Arnold networks (CP-KAN), a neural architecture that uses Chebyshev polynomials as the learnable univariate functions in each neuron. Its central claim is that choosing the right polynomial degree for every neuron—normally an exponential discrete search—can be rewritten as a quadratic unconstrained binary optimization (QUBO) problem and solved once per layer with simulated annealing. If true, this makes polynomial KANs practical, because the architecture can adapt its own complexity instead of requiring a fixed degree chosen by hand. The paper reports that on regression with limited data, CP-KAN reaches competitive accuracy with far fewer parameters than MLPs, recurrent networks, transformers, and other polynomial KAN baselines, and shows more stable training curves.

What carries the argument

The load-bearing device is the QUBO degree-selection objective of Eq. (7), whose fidelity term uses per-neuron least-squares costs $\mathrm{MSE}_{i,d}$ and whose penalty term enforces exactly one degree per neuron. Each neuron computes $f(x) = \sum_{i=0}^d c_i T_i(w^\top x + b)$, where $T_i$ are Chebyshev polynomials of the first kind generated by the recurrence $T_{n+1}(x)=2xT_n(x)-T_{n-1}(x)$, and the sum is cumulative up to the selected degree $d$. The QUBO formulation is what carries the argument: it converts an exponential combinatorial search into a quadratic binary program solvable by simulated annealing, and the resulting degree distribution is fixed before Phase 2 gradient training refines the continuous parameters. The coefficient-decay property of Chebyshev expansions for smooth functions is what justifies truncating at modest degrees.

What would settle it

Take a trained CP-KAN, freeze its weights, recompute $\mathrm{MSE}_{i,d}$ for every degree on the trained representations, and re-solve the QUBO. If the new degree set differs substantially from the Phase-1 set, and retraining from scratch with the new degrees gives no worse validation error, then the pre-training degree choice is not load-bearing. A comparison against random degree assignments of the same parameter count would isolate the QUBO's contribution.

Watch

Extended reading notes

Core claim

At the center of the paper is a constructive reduction. The discrete choice of a Chebyshev polynomial degree $d_i$ for each neuron $i$ is encoded with binary variables $q_{i,d}$, and the optimization objective is the QUBO in Eq. (7): minimize $\sum_{i,d} \mathrm{MSE}_{i,d} q_{i,d} + \alpha \sum_i (\sum_d q_{i,d}-1)^2$. The cost $\mathrm{MSE}_{i,d}$ is the per-neuron least-squares fit error of the cumulative Chebyshev transform $\Phi_{i,d}$ against the layer's target, so all degree costs are precomputable before any gradient training. Solving this QUBO by simulated annealing assigns a degree to every neuron in a layer in one step, lowering the search complexity from $O(D^N)$ to one optimization pass per layer. The trained network, after gradient fine-tuning, is reported to reach validation $R^2$ of 0.0376 on the financial market prediction benchmark with 1,571-2,547 parameters, compared with 0.0296 for a tuned MLP and negative $R^2$ for other polynomial KAN baselines, and a raw-scale house price MSE of 0.29 versus 1823 for an MLP. The paper also argues that Chebyshev bases are theoretically suited to mean-reverting financial dynamics, using the generator expansion for the Ornstein-Uhlenbeck process and a bound with exponential mean-reversion and $\log d / d$ approximation terms.

Load-bearing premise

The load-bearing premise is that degrees chosen from per-neuron least-squares fits before any joint training remain the right structure after the whole network is trained; if that independence fails, the QUBO step is an expensive heuristic.

Editorial extensions

If this is right

  • Degree selection in a polynomial KAN becomes a per-layer combinatorial optimization step, replacing $O(D^N)$ search with one QUBO solve; the paper reports near-linear scaling in network size and feasibility up to roughly 5,000 binary variables on CPU/GPU.
  • On the regression tasks tested, CP-KAN matches or beats MLPs, LSTMs, GRUs, and transformers with roughly an order of magnitude fewer parameters (1,571-2,547 versus 17,185-506,701), suggesting polynomial KANs can be parameter-efficient on tabular data.
  • Training stability is a reported advantage: CP-KAN maintains validation performance across extended epochs and learning rates from $10^{-5}$ to $10^{-2}$, while MLPs degrade on the financial task.
  • The theoretical link to mean-reverting processes gives a domain-specific justification: Chebyshev expansions approximate the Ornstein-Uhlenbeck generator with controlled error, so the architecture is a natural candidate for financial time-series modeling.
  • Classification remains a known limit: on the Covertype tabular benchmark, smooth polynomial bases are outperformed by tree ensembles, and image tasks only improve with width, so the paper's claim is specifically about regression.

Reading between the lines

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

  • Editorial extension: the same QUBO framing applies to any basis with precomputable per-neuron least-squares costs, so the degree-selection machinery is portable to Legendre, Hermite, or Fourier polynomial networks, not just Chebyshev.
  • Editorial extension: because degrees are fixed before training, alternating degree re-selection with gradient steps is a natural test of whether the optimal degree changes during training; if it does, iterative QUBO re-solving could improve the architecture.
  • Editorial extension: the paper's ablation of the plain Chebyshev KAN used a different parameter budget and a fixed degree, so the cleanest test of the QUBO's value is a fixed-degree Chebyshev KAN matched to CP-KAN's parameter count; that isolation is left implicit.
  • Editorial extension: the QUBO formulation is hardware-agnostic, so the same matrices solved by simulated annealing could be submitted to quantum annealers, making CP-KAN a candidate for near-term quantum-assisted architecture search.
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 / 5 minor

Summary. The paper introduces CP-KAN, a Kolmogorov-Arnold network whose activation functions are cumulative Chebyshev polynomials, and proposes to select per-neuron polynomial degrees via a QUBO formulation solved by simulated annealing. The claimed contribution is that this reduces degree assignment from O(D^N) to a single optimization step per layer. The authors also report regression experiments on financial and house-price data, classification studies on MNIST/CIFAR-10/Covertype, and a theoretical discussion connecting Chebyshev expansions to Ornstein-Uhlenbeck processes. The empirical sections show competitive regression performance with relatively few parameters, and the appendix contains substantial implementation detail and an available code repository.

Significance. If the degree-selection mechanism worked as advertised, the paper would offer a practical way to adapt polynomial complexity per neuron, with a concrete benefit in low-data regression. The manuscript also provides a broad empirical comparison, transparent hyperparameter tables, runtime measurements, and a public implementation, which are useful strengths. However, the central methodological claim is not supported by the written objective: as stated, the QUBO problem is separable and, without an explicit degree penalty, always selects the maximum degree. This inconsistency affects the core of the paper, not just the presentation. The theoretical link to mean reversion is also asserted rather than derived for the actual network. With those points addressed, the empirical findings could still be of interest, but in its current form the paper overclaims both the optimization contribution and the theoretical foundation.

major comments (4)
  1. [Sec. 3.2, Eq. (7); Appendix B; Fig. 6] The QUBO objective in Eq. (7) has no term coupling different neurons. The fidelity term is a sum over (i,d) of MSE_{i,d} q_{i,d}, and the one-hot penalty sum_i (sum_d q_{i,d} - 1)^2 couples degrees only within each neuron i. The problem therefore decomposes into N independent subproblems, each solvable by scanning degrees in O(D). Moreover, the Chebyshev basis in Eq. (6) is nested, so for any fixed projection w,b the least-squares training MSE_{i,d} is nonincreasing in d. Absent any explicit degree penalty, the minimizing assignment is q_{i,D}=1 for every neuron. This cannot produce the varied degree distributions shown in Fig. 6 and Appendix I.2. The hyperparameter tables list a 'Complexity Weight' (Tables 6-8 and 11) that is never defined in Eq. (7) or in Appendix B; this appears to be the undocumented mechanism responsible for the reported adaptive degrees. The authors must define the actual objective, including any degree-complexity term and its weighting, and show how the selected degrees are obtained from it.
  2. [Sec. 1 and Sec. 3.2; Abstract] The claimed complexity reduction from O(D^N) exhaustive search to 'a single optimization step per layer' is not supported. Since the stated objective is separable, the optimal solution is obtained by N independent O(D) comparisons after the per-neuron MSE costs are computed; no QUBO solver, simulated annealing, or combinatorial search is required. The O(D^N) baseline is therefore a strawman for the objective as written. If the authors intend a genuinely coupled problem, they need to introduce a coupling term (for example, a global degree budget or a layer-wise complexity constraint) and justify why the QUBO formulation is needed. If the problem is instead per-neuron degree selection, the combinatorial claims in the abstract and introduction should be removed and the contribution reframed accordingly.
  3. [Sec. 4.5 and Appendix D.2] The claim that CP-KAN shows 'robustness to input scales' and performs well on 'raw house prices' is contradicted by the preprocessing described in Appendix D.2, which states that the target is log-transformed with np.log1p() and that features are normalized with StandardScaler(). Table 1 reports 'Raw MSE' and 'Log MSE', but it is unclear what 'Raw' means given this preprocessing. The authors should either report results on genuinely unscaled targets and features, or revise the robustness claim to match the actual protocol. As written, the section overstates a result that the paper's own appendix does not support.
  4. [Sec. 5.3 and Appendix K] The theoretical justification is asserted rather than established. Theorem 1 in Appendix K states a standard generator decomposition for the Ornstein-Uhlenbeck SDE, but it is not applied to CP-KAN: no derivation shows that the CP-KAN architecture, its two-phase training, or its degree-selection rule inherits this expansion. Eq. (14) is presented as an error bound for the OU approximation, yet the constants, the norm, and the connection to the trained network are unspecified, and the sentence in the main text describes the bound in ways that do not follow from the appendix. The authors should either provide a formal statement linking the CP-KAN function class to these bounds, or present this material as heuristic motivation rather than 'theoretical analysis establishes connections'.
minor comments (5)
  1. [Eq. (6) and surrounding text] The degree variable d is used both as the maximum degree in Eq. (6) and as the summation bound, but the per-neuron degree d_i is not explicitly defined in the architecture section; distinguish the neuron-specific degree from the global maximum D.
  2. [Tables 6-8, 11] The 'Complexity Weight' hyperparameter is listed in multiple tables but is never defined in the objective; a short definition in Section 3.2 or Appendix B is needed for reproducibility.
  3. [Algorithm 1, Appendix C] The QUBO threshold T and the fixed fallback degree 3 are not specified with concrete values or a procedure for choosing them; state how T is set and how the default degree interacts with the reported degree distributions.
  4. [Fig. 3a and Appendix J] The phrase 'can be found here Fig. 3a' is a broken citation; fix the cross-reference and ensure all figure references in Sections 4.4 and 4.5 point to the correct subfigures.
  5. [Appendix D.1 and Sec. 4.4] The main text says experiments focus on a specific trading window, while Appendix D.1 describes a 70/30 split over 200,000 rows; clarify whether the trading window is the same as the data subset and state the temporal ordering of the split.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation; degree selection is fitting, not prediction, and empirical claims are externally benchmarked. Minor self-citation is not load-bearing.

full rationale

The paper's derivation chain is not circular. Eq. (7) transparently minimizes precomputed per-neuron least-squares MSE costs under a one-hot constraint; this is a model-fitting step, not a prediction, and the resulting degrees are subsequently used in a separate gradient-based training phase whose validation performance (Tables 1-2, Figs. 3-5) is measured against held-out data and external baselines. No claimed prediction is statistically forced by the degree-selection fit. The theoretical justification in Appendix K is imported from external sources (Applebaum 2007; Burtnyak and Malyts'ka 2018), not from a self-citation chain. The self-citations to QKAN (Ivashkov et al. 2024) and the implementation repository are contextual and do not carry the central claim. Two non-circular technical flaws should be flagged: (i) Eq. (7) is separable over neurons, so its QUBO optimum is just the per-neuron argmin of MSE_{i,d}; because the Chebyshev basis is cumulative, the unpenalized objective would always select the maximal degree, and the varied degree distributions in Fig. 6 and Appendix I.2 imply an undocumented complexity penalty — the 'Complexity Weight' listed in Tables 6-8 and 11 is never defined in the objective. (ii) The O(D^N) exhaustive-search baseline is a strawman for a separable objective. These are reproducibility/correctness concerns, not circular reasoning, so the circularity score remains low. The untested assumption that Phase-1 degree choices remain appropriate after Phase-2 gradient training is likewise a limitation, not a circular step.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central method rests on a handful of free hyperparameters and a key modeling assumption that per-neuron independent least-squares fits guide the final network structure. The theoretical financial justification is an externally borrowed OU-process result, not a derivation for CP-KAN. No new physical or conceptual entities are introduced; the architecture reuses existing Chebyshev bases and QUBO optimization.

free parameters (4)
  • QUBO penalty weight alpha
    Controls the one-hot constraint in Eq. 7; no value is reported in the paper, so it is a hand-tuned hyperparameter that affects degree feasibility.
  • Complexity weight = searched over {0.05, 0.1, 0.2}
    Used in grid searches to regularize degree selection, but not formally defined in the main text.
  • Maximum polynomial degree D = dataset-dependent from {3,5,7,9,11}
    Defines the search space for degree selection and is chosen per dataset.
  • Default hidden degree = 4 in main text, 3 to 5 in experiments
    Degree assigned when QUBO is skipped for large layers; a manual fallback.
assumptions (5)
  • standard math Any continuous multivariate function can be represented via the Kolmogorov-Arnold theorem (Eq. 1).
    Justifies the KAN architecture, but the theorem's representation guarantees nothing about learnability of the specific Chebyshev parameterization.
  • standard math Chebyshev polynomials form a complete orthogonal basis on [-1,1] with favorable approximation properties.
    Basis for the activation functions in Eq. 5 and 6.
  • domain assumption Financial time series can be modeled as locally stationary, mean-reverting processes whose generators admit Chebyshev expansions.
    Underpins the theoretical justification in Sec. 5.3 and Appendix K; never validated on the Jane Street data.
  • ad hoc to paper Per-neuron least-squares MSE computed independently is a valid proxy for the final multi-layer network loss when selecting degrees.
    Used to compute MSE_{i,d} in Eq. 7 and Phase 1; no evidence that layer-wise independent degree choices remain optimal after joint training.
  • domain assumption Simulated annealing on the QUBO reliably finds a good one-hot degree assignment.
    Phase 1 relies on this heuristic; no optimality guarantee or comparison to exact solvers on large instances.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Degree-Optimized Cumulative Polynomial Kolmogorov-Arnold Networks." pith.science (2026). https://pith.science/paper/CASYMG75

@misc{pith2026250515228,
  author       = {Pith},
  title        = {Pith review of: Degree-Optimized Cumulative Polynomial Kolmogorov-Arnold Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CASYMG75}},
  note         = {Machine review of arXiv:2505.15228}
}
abstract

We introduce cumulative polynomial Kolmogorov-Arnold networks (CP-KAN), a neural architecture combining Chebyshev polynomial basis functions and quadratic unconstrained binary optimization (QUBO). Our primary contribution involves reformulating the degree selection problem as a QUBO task, reducing the complexity from $O(D^N)$ to a single optimization step per layer. This approach enables efficient degree selection across neurons while maintaining computational tractability. The architecture performs well in regression tasks with limited data, showing good robustness to input scales and natural regularization properties from its polynomial basis. Additionally, theoretical analysis establishes connections between CP-KAN's performance and properties of financial time series. Our empirical validation across multiple domains demonstrates competitive performance compared to several traditional architectures tested, especially in scenarios where data efficiency and numerical stability are important. Our implementation, including strategies for managing computational overhead in larger networks is available in Ref.~\citep{cpkan_implementation}.

Figures

Figures reproduced from arXiv: 2505.15228 by the authors.

Figure 1
Figure 1. Single layer architecture of CP-KAN showing input features ( [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Performance comparison across KAN architectures on the Jane Street Market Prediction [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Comparison of training behavior and generalization between CP-KAN and MLPs across [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Training stability comparison between MLP and CP-KAN on the Jane Street Market [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Impact of model complexity on MNIST test accuracy for CP-KAN solved via QUBO [PITH_FULL_IMAGE:figures/full_fig_p021_5.png]
Figure 6
Figure 6. Figure 6: Comparison of polynomial degree distributions selected by different optimization methods [PITH_FULL_IMAGE:figures/full_fig_p022_6.png]
Figure 7
Figure 7. Figure 7: Log-log plot illustrating the computational scaling of the QUBO optimization time versus [PITH_FULL_IMAGE:figures/full_fig_p022_7.png]
Figure 8
Figure 8. Figure 8: Impact of model complexity on CIFAR-10 test accuracy for CP-KAN optimized with [PITH_FULL_IMAGE:figures/full_fig_p023_8.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

25 extracted references · 14 canonical work pages

  1. [5]

    The efficiency of Chebyshev polynomials stems from their approximation properties for smooth functions. Assuming a function g(x) defined on [−1, 1] has k continuous derivatives, its Cheby- shev expansion g(x) ≈ P∞ i=0 ciTi(x) exhibits coefficients ci that decay algebraically, typically as O(i−(k+1)) [Majidian, 2017]. This rapid decay ensures that smooth f...

    work page 2017

  2. [10]

    URL https://arxiv.org/abs/2105.08806. A. G. Ivakhnenko. Polynomial theory of complex systems. IEEE Transactions on Systems, Man, and Cybernetics, SMC-1(4):364–378,

    work page Pith review arXiv

  3. [13]

    doi: 10.3390/e24030348

    ISSN 1099-4300. doi: 10.3390/e24030348. URL https://www.mdpi.com/1099-4300/24/3/348. Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. Deep learning. Nature, 521(7553):436–444,

    work page doi:10.3390/e24030348

  4. [15]

    Quantum-inspired activation functions and quantum Chebyshev-polynomial network

    URL https://arxiv.org/abs/ 2404.05901. Hangyu Liu, Gabriel Bender, Yichen Sun, Alvin Wan, Ruihan Yang, and Kevin Swersky. Gradient- free neural architecture search via zero-cost proxies. arXiv preprint arXiv:2301.05787,

    work page Pith review arXiv

  5. [17]

    Hrushikesh Narhar Mhaskar and Devidas V

    URL https://arxiv.org/abs/2305.02997. Hrushikesh Narhar Mhaskar and Devidas V . Pai. Fundamentals of Approximation Theory. CRC Press,

    arXiv

  6. [20]

    Seyd Teymoor Seydi

    URL https://arxiv.org/ abs/2501.03540. Seyd Teymoor Seydi. Exploring the potential of polynomial basis functions in kolmogorov-arnold networks: A comparative study of different groups of polynomials,

    arXiv

  7. [21]

    org/abs/2406.02583

    URL https://arxiv. org/abs/2406.02583. S.S. Sidharth, A.R. Keerthana, R. Gokul, and K.P. Anas. Chebyshev polynomial-based Kolmogorov- Arnold networks: An efficient architecture for nonlinear function approximation,

    arXiv

  8. [23]

    URL https://arxiv.org/abs/2405.07200. Lloyd N. Trefethen. Approximation Theory and Approximation Practice. Society for Industrial and Applied Mathematics, Philadelphia, PA,

    arXiv

Show all 25 references
  1. [24]

    time step

    (Fixed) Original-KAN (Spline) Hidden Dim {16, 20, 24, 32} Num (Spline Intervals) {3, 5, 7} K (Spline Order) {2, 3, 4} Wavelet-KAN Hidden Dim {16, 20, 24, 32} With Batch Norm {True, False} Wavelet Type ’mexican hat’ (Fixed) Jacobi-KAN Hidden Dim {16, 20, 24, 32} Degree {3, 5, 7...

    2021

  2. [1957]

    Vladimir Arnold

    URL https://doi.org/10.1007/978-3-642-01742-1_2 . Vladimir Arnold. On the representation of continuous functions of three variables by superpositions of continuous functions of two variables. Mat. Sb. (N.S.), 48(90)(1):3–74,

    doi

  3. [1959]

    CP-KAN Authors

    URL https: //doi.org/10.1007/978-3-642-01742-1_6 . CP-KAN Authors. CP-KAN Implementation. https://github.com/Mathewvanh/CP-KAN,

    doi

  4. [1983]

    URL https://doi.org/10.1126/ science.220.4598.671

    doi: 10.1126/science.220.4598.671. URL https://doi.org/10.1126/ science.220.4598.671. Andrey Kolmogorov. On the representation of continuous functions of several variables by superpo- sitions of continuous functions of a smaller number of variables. Dokl. Akad. Nauk SSSR, 108:...

    doi

  5. [1989]

    D-Wave Systems Inc

    URL https://doi.org/10.1007/BF02551274. D-Wave Systems Inc. Neal: A python module for optimizing binary quadratic models using simulated annealing,

    doi

  6. [1997]

    Shentong Mo, Jingfei Xia, and Pinxu Ren

    ISBN 9780070428072. Shentong Mo, Jingfei Xia, and Pinxu Ren. Simulated annealing for neural architecture search. In Workshop on Optimization for Machine Learning (OPT) at NeurIPS 2021, Pittsburgh, PA, USA,

    2021

  7. [2001]

    Karen Simonyan and Andrew Zisserman

    doi: 10.1109/4235.910465. Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556,

  8. [2006]

    ISBN 978-0-387-31073-2. Z. Bozorgasl and H. Chen. Wav-kan: Wavelet kolmogorov-arnold networks. arXiv preprint arXiv:2405.12832,

    arXiv

  9. [2015]

    Shaozhi Li, M Sabbir Salek, Yao Wang, and Mashrur Chowdhury

    URL https://doi.org/10.1038/nature14539. Shaozhi Li, M Sabbir Salek, Yao Wang, and Mashrur Chowdhury. Quantum-inspired activation functions and quantum chebyshev-polynomial network,

    doi

  10. [2016]

    Jingwei Hu, Xiaodong Huang, Jie Shen, and Haizhao Yang

    URL https://doi.org/10.1109/CVPR.2016.90. Jingwei Hu, Xiaodong Huang, Jie Shen, and Haizhao Yang. A fast petrov-galerkin spectral method for the multi-dimensional boltzmann equation using mapped chebyshev functions,

    2016 doi

  11. [2017]

    11 Duncan McElfresh, Sujay Khandagale, Jonathan Valverde, Vishak Prasad C, Benjamin Feuer, Chin- may Hegde, Ganesh Ramakrishnan, Micah Goldblum, and Colin White

    doi: 10.1016/j.apnum.2016.11.004. 11 Duncan McElfresh, Sujay Khandagale, Jonathan Valverde, Vishak Prasad C, Benjamin Feuer, Chin- may Hegde, Ganesh Ramakrishnan, Micah Goldblum, and Colin White. When do neural nets outperform boosted trees on tabular data?,

    2016 doi

  12. [2019]

    Weijieying Ren, Tianxiang Zhao, Yuqing Huang, and Vasant Honavar

    URL https: //arxiv.org/abs/1912.01703. Weijieying Ren, Tianxiang Zhao, Yuqing Huang, and Vasant Honavar. Deep learning within tabular data: Foundations, challenges, advances and future directions,

    1912 arXiv

  13. [2020]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pages 770–778,

    2016

  14. [2021]

    doi: 10.22331/ q-2021-04-08-428

    ISSN 2521-327X. doi: 10.22331/ q-2021-04-08-428. URL https://doi.org/10.22331/q-2021-04-08-428 . Ian Goodfellow, Yoshua Bengio, and Aaron Courville. Deep Learning. MIT Press,

    2021 doi

  15. [2022]

    Xiao Han, Xinfeng Zhang, Yiling Wu, Zhenduo Zhang, and Zhe Wu

    URL https://proceedings.neurips.cc/paper_files/paper/2022/file/ 0378c7692da36807bdec87ab043cdadc-Paper-Datasets_and_Benchmarks.pdf. Xiao Han, Xinfeng Zhang, Yiling Wu, Zhenduo Zhang, and Zhe Wu. Kan4tsf: Are kan and kan-based models effective for time series forecasting? In arXiv,

    2022

  16. [2024]

    Diederik P

    URL https://arxiv.org/abs/2410.04435. Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In3rd International Conference on Learning Representations (ICLR),

    arXiv

  17. [2025]

    Christopher M

    Accessed: 2025-05-18. Christopher M. Bishop. Pattern Recognition and Machine Learning (Information Science and Statistics). Springer-Verlag, Berlin, Heidelberg,

    2025

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.