pith. machine review for the scientific record. sign in

arxiv: 2604.18414 · v1 · submitted 2026-04-20 · 💻 cs.LG · cs.NA· math.NA

Recognition: unknown

Balance-Guided Sparse Identification of Multiscale Nonlinear PDEs with Small-coefficient Terms

Zhenhua Dang , Lei Zhang , Long Wang , Guowei He
Authors on Pith no claims yet

Pith reviewed 2026-05-10 04:52 UTC · model grok-4.3

classification 💻 cs.LG cs.NAmath.NA
keywords sparse identificationgoverning equationsmultiscale PDEssmall coefficientsdominant balancedata-driven discoverynonlinear systemsSINDy
0
0 comments X

The pith

Balance-guided sparse regression identifies small-coefficient terms in nonlinear PDEs by their contribution to the equation balance.

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

The paper introduces Balance-Guided SINDy to solve the problem that standard sparse regression discards important terms when their coefficients are small in multiscale systems. It ranks candidate terms by how much each one contributes to the overall balance of the governing equation at every point, then repeatedly fits the current model by least squares and drops only the negligible contributors. This preserves dynamically significant terms even when their sizes are orders of magnitude smaller than others. A reader would care because many physical systems, from fluid waves to chemical reactions, contain such small but essential terms, and missing them produces incorrect predictions from data.

Core claim

Balance-Guided SINDy reformulates the sparse regression task as a term-level ℓ_{2,0}-regularized problem and solves it with a progressive pruning strategy. Terms are ranked according to their relative contributions to the governing equation balance rather than their absolute coefficient magnitudes. By alternating between least-squares regression and elimination of negligible terms, the method retains dynamically significant terms even when their coefficients are small. Experiments on the Korteweg-de Vries equation with a small dispersion coefficient, a modified Burgers equation with vanishing hyperviscosity, a modified Kuramoto-Sivashinsky equation with multiple small-coefficient terms, and

What carries the argument

The progressive pruning strategy that ranks and eliminates terms according to their relative contribution to the instantaneous governing-equation balance.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same balance-ranking idea could be tested on ordinary differential equations or on real experimental time series from engineering systems.
  • If the ranking proves stable, it might be combined with other sparsity penalties to handle even higher-dimensional or chaotic multiscale problems.
  • A direct check would be to apply the method to data generated from a known PDE whose small term is known to control long-term behavior and verify recovery.

Load-bearing premise

Ranking terms by relative contribution to the instantaneous governing-equation balance correctly identifies which small-coefficient terms are dynamically significant.

What would settle it

An experiment on the Korteweg-de Vries equation in which the dispersion coefficient is made progressively smaller while noise is added to the data, showing whether the small term is still recovered or is incorrectly pruned.

Figures

Figures reproduced from arXiv: 2604.18414 by Guowei He, Lei Zhang, Long Wang, Zhenhua Dang.

Figure 1
Figure 1. Figure 1: Schematic of the BG-SINDy method. (a) Spatio-temporal data sampling and [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Results for the KdV equation. (a) History of residual in the sparsification pro [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Heatmap of the coefficient error in the BG-SINDy discovery of the KdV equation [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Results for the modified Burgers equation. (a) History of residual in the spar [PITH_FULL_IMAGE:figures/full_fig_p017_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Results for the modified KS equation. (a) History of residual in the sparsification [PITH_FULL_IMAGE:figures/full_fig_p019_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Results for the 2D reaction–diffusion equation ( [PITH_FULL_IMAGE:figures/full_fig_p021_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Results for the 2D reaction–diffusion equation ( [PITH_FULL_IMAGE:figures/full_fig_p022_7.png] view at source ↗
read the original abstract

Data-driven discovery of governing equations has advanced significantly in recent years; however, existing methods often struggle in multiscale systems where dynamically significant terms may have small coefficients. Therefore, we propose Balance-Guided SINDy (BG-SINDy) inspired by the principle of dominant balance, which reformulates $\ell_0$-constrained sparse regression as a term-level $\ell_{2,0}$-regularized problem and solves it using a progressive pruning strategy. Terms are ranked according to their relative contributions to the governing equation balance rather than their absolute coefficient magnitudes. Based on this criterion, BG-SINDy alternates between least-squares regression and elimination of negligible terms, thereby preserving dynamically significant terms even when their coefficients are small. Numerical experiments on the Korteweg--de Vries equation with a small dispersion coefficient, a modified Burgers equation with vanishing hyperviscosity, a modified Kuramoto--Sivashinsky equation with multiple small-coefficient terms, and a two-dimensional reaction--diffusion system demonstrate the validity of BG-SINDy in discovering small-coefficient terms. The proposed method thus provides an efficient approach for discovering governing equations that contain small-coefficient terms.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper introduces Balance-Guided SINDy (BG-SINDy), which reformulates ℓ₀-constrained sparse regression for PDE discovery as a term-level ℓ_{2,0} problem solved by progressive pruning. Terms are ranked and eliminated according to their relative contribution to the instantaneous residual balance of the governing equation rather than absolute coefficient size. This is claimed to recover dynamically significant small-coefficient terms in multiscale systems. Validity is asserted via numerical experiments on the KdV equation with small dispersion, a modified Burgers equation, a modified Kuramoto–Sivashinsky equation, and a 2D reaction–diffusion system.

Significance. If the central claim holds, the work would supply a lightweight, balance-aware extension to the SINDy family that addresses a recurring practical failure mode—loss of small but dynamically essential terms—without adding free parameters beyond the pruning schedule. The procedural nature (alternating least-squares and balance-ranked elimination) is attractive for reproducibility, yet the absence of quantitative recovery metrics, ablation studies, and out-of-sample dynamical tests limits the strength of the significance assessment at present.

major comments (3)
  1. [§4] §4 (numerical experiments): The four example systems are presented only with qualitative recovery statements; no coefficient error norms, trajectory prediction errors on held-out initial conditions, or long-time integration tests are reported. This leaves the central claim that small-coefficient terms are “preserved” without quantitative support and prevents direct comparison to standard SINDy or other balance-aware baselines.
  2. [§3] §3 (algorithm description): No derivation or invariance argument is supplied showing that the relative-balance ranking remains reliable when the small term is invisible in the training trajectory or becomes dominant only after long-time evolution. The skeptic’s concern is therefore not addressed: the method could still discard a term whose coefficient is orders of magnitude smaller yet essential outside the sampled regime.
  3. [Table 1 / Figure 2] Table 1 / Figure 2 (KdV and modified Burgers results): The reported “recovered” equations are shown without the corresponding residual norms or comparison against a simple coefficient-threshold baseline run on the same data; it is therefore impossible to isolate the contribution of the balance-guided pruning step.
minor comments (2)
  1. [§3] Notation for the balance residual and the pruning threshold schedule is introduced without a compact algorithmic pseudocode block; a single-line algorithm box would improve clarity.
  2. [Abstract] The abstract states that the method “demonstrates the validity” of BG-SINDy, yet the experiments section contains no statistical summary across multiple noise realizations or initial-condition ensembles.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for the constructive and insightful comments. We address each major point below, indicating the revisions we will incorporate to strengthen the manuscript.

read point-by-point responses

  1. Referee: [§4] §4 (numerical experiments): The four example systems are presented only with qualitative recovery statements; no coefficient error norms, trajectory prediction errors on held-out initial conditions, or long-time integration tests are reported. This leaves the central claim that small-coefficient terms are “preserved” without quantitative support and prevents direct comparison to standard SINDy or other balance-aware baselines.

    Authors: We agree that quantitative metrics are necessary to rigorously support the central claims and enable comparisons. In the revised manuscript we will add relative coefficient error norms, out-of-sample trajectory prediction errors on held-out initial conditions, and long-time integration stability tests for all four example systems. These additions will also include direct comparisons against standard SINDy. revision: yes

  2. Referee: [§3] §3 (algorithm description): No derivation or invariance argument is supplied showing that the relative-balance ranking remains reliable when the small term is invisible in the training trajectory or becomes dominant only after long-time evolution. The skeptic’s concern is therefore not addressed: the method could still discard a term whose coefficient is orders of magnitude smaller yet essential outside the sampled regime.

    Authors: This concern is valid. The balance-ranking criterion is computed from the training data; if a term contributes negligibly to the observed residual balance it may be pruned even if it becomes important later. We will expand the discussion in §3 to explicitly state this assumption and the requirement for sufficiently rich training trajectories, but a full invariance proof lies beyond the current scope. revision: partial

  3. Referee: [Table 1 / Figure 2] Table 1 / Figure 2 (KdV and modified Burgers results): The reported “recovered” equations are shown without the corresponding residual norms or comparison against a simple coefficient-threshold baseline run on the same data; it is therefore impossible to isolate the contribution of the balance-guided pruning step.

    Authors: We will augment Table 1 and Figure 2 with residual norms of the recovered equations and add side-by-side comparisons against a standard coefficient-threshold SINDy baseline applied to identical data. This will isolate the effect of the balance-guided pruning. revision: yes

standing simulated objections not resolved
  • A rigorous derivation or invariance argument guaranteeing that relative-balance ranking recovers terms invisible in the training data but dominant at long times.

Circularity Check

0 steps flagged

No significant circularity: BG-SINDy is a procedural heuristic on standard sparse regression

full rationale

The paper introduces BG-SINDy as an algorithmic modification to SINDy that ranks candidate terms by their relative contribution to the instantaneous residual balance and prunes via progressive elimination before re-solving least squares. This construction does not reduce any claimed result to a fitted parameter that is later renamed a prediction, nor does it rely on a self-citation chain or uniqueness theorem imported from the authors' prior work. The governing equations recovered in the numerical examples are obtained directly from the data via the described procedure; they are not equivalent to the inputs by definition. The central modeling choice (balance-based ranking) is presented as an empirical heuristic inspired by dominant balance, not as a derived identity. Hence the derivation chain is self-contained and non-circular.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The method rests on the domain assumption that dominant balance can be evaluated pointwise from data and that relative balance contribution is a reliable proxy for dynamical importance. No new physical entities are postulated.

free parameters (1)
  • pruning threshold schedule
    The progressive elimination step requires at least one tunable tolerance or schedule that determines when a term is deemed negligible.
axioms (1)
  • domain assumption The principle of dominant balance holds for the target multiscale systems.
    The ranking criterion is explicitly inspired by dominant balance.

pith-pipeline@v0.9.0 · 5518 in / 1263 out tokens · 27183 ms · 2026-05-10T04:52:25.481675+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

70 extracted references · 44 canonical work pages · 1 internal anchor

  1. [1]

    H. Voss, M. J. Bünner, M. Abel, Identification of continuous, spatiotem- poral systems, Phys. Rev. E 57 (1998) 2820–2823. https://doi.org/ 10.1103/PhysRevE.57.2820

    work page doi:10.1103/physreve.57.2820 1998

  2. [2]

    M. Bär, R. Hegger, H. Kantz, Fitting partial differential equations to space-time dynamics, Phys. Rev. E 59 (1999) 337–342. https://doi. org/10.1103/PhysRevE.59.337

    work page doi:10.1103/physreve.59.337 1999

  3. [3]

    Schmidt, H

    M. Schmidt, H. Lipson, Distilling Free-Form Natural Laws from Exper- imental Data, Science 324 (5923) (2009) 81–85. https://doi.org/10. 1126/science.1165893

    2009

  4. [4]

    Bongard, H

    J. Bongard, H. Lipson, Automated reverse engineering of nonlinear dy- namical systems, Proc. Natl. Acad. Sci. U.S.A. 104 (24) (2007) 9943–

    2007

  5. [5]

    https://doi.org/10.1073/pnas.0609476104

    work page doi:10.1073/pnas.0609476104

  6. [6]

    S. L. Brunton, J. L. Proctor, J. N. Kutz, Discovering governing equations from data by sparse identification of nonlinear dynamical systems, Proc. Natl. Acad. Sci. U.S.A. 113 (15) (2016) 3932–3937. https://doi.org/ 10.1073/pnas.1517384113

    work page doi:10.1073/pnas.1517384113 2016

  7. [7]

    S. H. Rudy, S. L. Brunton, J. L. Proctor, J. N. Kutz, Data-driven dis- covery of partial differential equations, Sci. Adv. 3 (4) (2017) e1602614. https://doi.org/10.1126/sciadv.1602614

    work page doi:10.1126/sciadv.1602614 2017

  8. [8]

    GlobularClusterAges,

    K. Champion, B. Lusch, J. N. Kutz, S. L. Brunton, Data-driven dis- covery of coordinates and governing equations, Proc. Natl. Acad. Sci. U.S.A. 116 (45) (2019) 22445–22451. https://doi.org/10.1073/pnas. 1906995116

    work page doi:10.1073/pnas 2019

  9. [9]

    L. Fung, U. Fasel, M. Juniper, Rapid Bayesian identification of sparse nonlinear dynamics from scarce and noisy data, Proc. R. Soc. A, Math. Phys. Eng. Sci. 481 (2307) (2025) 20240200. https://doi.org/10. 1098/rspa.2024.0200

    work page arXiv 2025

  10. [10]

    Prandtl, über flüssigkeitsbewegung bei sehr kleiner reibung, in: Ver- handlungen des III

    L. Prandtl, über flüssigkeitsbewegung bei sehr kleiner reibung, in: Ver- handlungen des III. Internationalen Mathematiker-Kongresses, Heidel- berg 1904, B. G. Teubner, Leipzig, 1905, pp. 484–491. 23

    1904

  11. [11]

    Schlichting, K

    H. Schlichting, K. Gersten, Boundary-Layer Theory, 9th Edi- tion, Springer, Berlin, 2017. https://doi.org/10.1007/ 978-3-662-52919-5

    2017

  12. [12]

    C. D. Murray, S. F. Dermott, Solar System Dynamics, Cambridge University Press, Cambridge, 2000. https://doi.org/10.1017/ CBO9781139174817

    2000

  13. [13]

    G.-J. Both, S. Choudhury, P. Sens, R. Kusters, DeepMoD: Deep learning for model discovery in noisy data, J. Comput. Phys. 428 (2021) 109985. https://doi.org/10.1016/j.jcp.2020.109985

    work page doi:10.1016/j.jcp.2020.109985 2021

  14. [14]

    C. M. Bender, S. A. Orszag, Advanced mathematical methods for scientists and engineers: Asymptotic methods and perturba- tion theory, Springer, New York, 1999. https://doi.org/10.1007/ 978-1-4757-3069-2

    1999

  15. [15]

    A. H. Nayfeh, Perturbation methods, John Wiley & Sons, Hoboken, 2024

    2024

  16. [16]

    M. H. Holmes, Introduction to perturbation methods, Springer, New York, 2012

    2012

  17. [17]

    J. K. Kevorkian, J. D. Cole, Multiple scale and singular perturbation methods, Vol. 114, Springer, New York, 2012

    2012

  18. [18]

    D. A. Messenger, D. M. Bortz, Weak SINDy for partial differential equations, J. Comput. Phys. 443 (2021) 110525. https://doi.org/ 10.1016/j.jcp.2021.110525

    work page doi:10.1016/j.jcp.2021.110525 2021

  19. [19]

    Hance and Sabine Hossenfelder

    U. Fasel, J. N. Kutz, B. W. Brunton, S. L. Brunton, Ensemble-SINDy: Robust sparse model discovery in the low-data, high-noise limit, with active learning and control, Proc. R. Soc. A, Math. Phys. Eng. Sci. 478 (2260) (2022) 20210904. https://doi.org/10.1098/rspa.2021. 0904

    work page doi:10.1098/rspa.2021 2022

  20. [20]

    B. M. de Silva, K. Champion, M. Quade, J.-C. Loiseau, J. N. Kutz, S. L. Brunton, PySINDy: A Python package for the sparse identification of nonlinear dynamical systems from data, J. Open Source Softw. 5 (49) (2020) 2104. https://doi.org/10.21105/joss.02104. 24

    work page doi:10.21105/joss.02104 2020

  21. [21]

    Zhang, S

    L. Zhang, S. Tang, G. He, Learning chaotic systems from noisy data via multi-step optimization and adaptive training, Chaos 32 (12) (2022) 123134. https://doi.org/10.1063/5.0114542

    work page doi:10.1063/5.0114542 2022

  22. [22]

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations

    M. Raissi, P. Perdikaris, G. Karniadakis, Physics-informed neural net- works: A deep learning framework for solving forward and inverse prob- lems involving nonlinear partial differential equations, J. Comput. Phys. 378 (2019) 686–707. https://doi.org/10.1016/j.jcp.2018.10.045

    work page doi:10.1016/j.jcp.2018.10.045 2019

  23. [23]

    G. E. Karniadakis, I. G. Kevrekidis, L. Lu, P. Perdikaris, S. Wang, L. Yang, Physics-informed machine learning, Nat. Rev. Phys. 3 (6) (2021) 422–440. https://doi.org/10.1038/s42254-021-00314-5

    work page doi:10.1038/s42254-021-00314-5 2021

  24. [24]

    Z. Chen, Y. Liu, H. Sun, Physics-informed learning of governing equa- tions from scarce data, Nat. Commun. 12 (1) (2021) 6136. https: //doi.org/10.1038/s41467-021-26434-1

    work page doi:10.1038/s41467-021-26434-1 2021

  25. [25]

    Zhang, G

    L. Zhang, G. He, Scale-Decomposed Physics-Informed Neural Networks for singular perturbation problems, Commun. Comput. Phys. 39 (4) (2026) 1267–1298. https://doi.org/10.4208/cicp.OA-2024-0149

    work page doi:10.4208/cicp.oa-2024-0149 2026

  26. [26]

    Science Advances6(16), eaay2631 (2020).https: //doi.org/10.1126/sciadv.aay2631

    S.-M. Udrescu, M. Tegmark, AI Feynman: A physics-inspired method for symbolic regression, Sci. Adv. 6 (16) (2020) eaay2631. https:// doi.org/10.1126/sciadv.aay2631

    work page doi:10.1126/sciadv.aay2631 2020

  27. [27]

    Y. Chen, Y. Luo, Q. Liu, H. Xu, D. Zhang, Symbolic genetic algo- rithm for discovering open-form partial differential equations (SGA- PDE), Phys. Rev. Res. 4 (2) (2022) 023174. https://doi.org/10. 1103/PhysRevResearch.4.023174

    2022

  28. [28]

    Interpretable Machine Learning for Science with PySR and SymbolicRegression.jl

    M. Cranmer, Interpretable Machine Learning for Science with PySR and SymbolicRegression.jl (2023). arXiv:2305.01582

    work page internal anchor Pith review arXiv 2023

  29. [29]

    Z. Liu, Y. Wang, S. Vaidya, F. Ruehle, J. Halverson, M. Soljacic, T. Y. Hou, M. Tegmark, KAN: Kolmogorov–Arnold networks, in: The Thir- teenth International Conference on Learning Representations, 2025

    2025

  30. [30]

    Z. Zhou, Z. Xu, Y. Liu, S. Wang, asKAN: Active subspace embedded Kolmogorov–Arnold network, Neural Netw. 195 (2026) 108280. https: //doi.org/10.1016/j.neunet.2025.108280. 25

    work page doi:10.1016/j.neunet.2025.108280 2026

  31. [31]

    M. Du, Y. Chen, Z. Wang, L. Nie, D. Zhang, Large language models for automatic equation discovery of nonlinear dynamics, Phys. Fluids 36 (9) (2024) 097121. https://doi.org/10.1063/5.0224297

    work page doi:10.1063/5.0224297 2024

  32. [32]

    H. Xu, Y. Chen, R. Cao, T. Tang, M. Du, J. Li, A. H. Callaghan, D. Zhang, Generative discovery of partial differential equations by learning from math handbooks, Nat. Commun. 16 (1) (2025) 10255. https://doi.org/10.1038/s41467-025-65114-2

    work page doi:10.1038/s41467-025-65114-2 2025

  33. [33]

    D. R. Gurevich, M. R. Golden, P. A. Reinbold, R. O. Grigoriev, Learning fluid physics from highly turbulent data using sparse physics-informed discovery of empirical relations (SPIDER), J. Fluid Mech. 996 (2024) A25. https://doi.org/10.1017/jfm.2024.813

    work page doi:10.1017/jfm.2024.813 2024

  34. [34]

    C. J. Wareing, A. T. Roy, M. Golden, R. O. Grigoriev, S. M. Tobias, Data-driven discovery of the equations of turbulent convection, Geophys. Astrophys. Fluid Dyn. 119 (3–4) (2025) 550–582. https://doi.org/10. 1080/03091929.2025.2509469

    work page arXiv 2025

  35. [35]

    Golden, R

    M. Golden, R. O. Grigoriev, J. Nambisan, A. Fernandez-Nieves, Physi- cally informed data-driven modeling of active nematics, Sci. Adv. 9 (27) (2023) eabq6120. https://doi.org/10.1126/sciadv.abq6120

    work page doi:10.1126/sciadv.abq6120 2023

  36. [36]

    Golden, K

    M. Golden, K. Satapathy, D. Psaltis, Scalable Discovery of Fundamental Physical Laws: Learning Magnetohydrodynamics from 3D Turbulence Data (2025). arXiv:2501.04094

    work page arXiv 2025

  37. [37]

    Golden, Scalable Sparse Regression for Model Discovery: The Fast Lane to Insight (2024)

    M. Golden, Scalable Sparse Regression for Model Discovery: The Fast Lane to Insight (2024). arXiv:2405.09579

    work page arXiv 2024

  38. [38]

    Z. Li, H. Yuan, W. Han, Y. Hou, H. Li, H. Ding, Z. Jiang, L. Yang, Bi- level identification of governing equations for nonlinear physical systems, Nat. Comput. Sci. 5 (6) (2025) 456–466. https://doi.org/10.1038/ s43588-025-00804-x

    2025

  39. [39]

    E. J. Hinch, Perturbation Methods, Cambridge University Press, Cam- bridge, 1991. https://doi.org/10.1017/CBO9781139172189

    work page doi:10.1017/cbo9781139172189 1991

  40. [40]

    Kevorkian, J

    J. Kevorkian, J. D. Cole, Multiple Scale and Singular Perturbation Methods, Springer, New York, 1996. https://doi.org/10.1007/ 978-1-4612-3968-0 . 26

    1996

  41. [41]

    J. L. Callaham, J. V. Koch, B. W. Brunton, J. N. Kutz, S. L. Brunton, Learning dominant physical processes with data-driven balance mod- els, Nat. Commun. 12 (1) (2021) 1016. https://doi.org/10.1038/ s41467-021-21331-z

    2021

  42. [42]

    Fukami, T

    K. Fukami, T. Murata, K. Zhang, K. Fukagata, Sparse identification of nonlinear dynamics with low-dimensionalized flow representations, J. Fluid Mech. 926 (2021) A10. https://doi.org/10.1017/jfm.2021. 697

    work page doi:10.1017/jfm.2021 2021

  43. [43]

    Kaheman, J

    K. Kaheman, J. N. Kutz, S. L. Brunton, SINDy-PI: a robust algorithm for parallel implicit sparse identification of nonlinear dynamics, Proc. R. Soc. A, Math. Phys. Eng. Sci. 476 (2234) (2020) 20200279. https: //doi.org/10.1098/rspa.2020.0279

    work page doi:10.1098/rspa.2020.0279 2020

  44. [44]

    A. G. Baydin, B. A. Pearlmutter, A. A. Radul, J. M. Siskind, Automatic Differentiation in Machine Learning: a Survey, J. Mach. Learn. Res. 18 (153) (2018) 1–43

    2018

  45. [45]

    H. Xu, H. Chang, D. Zhang, DL-PDE: Deep-learning based data-driven discovery of partial differential equations from discrete and noisy data, Commun. Comput. Phys. 29 (3) (2021) 698–728. https://doi.org/ 10.4208/cicp.OA-2020-0142

    work page doi:10.4208/cicp.oa-2020-0142 2021

  46. [46]

    Kaheman, S

    K. Kaheman, S. L. Brunton, J. Nathan Kutz, Automatic differentiation to simultaneously identify nonlinear dynamics and extract noise proba- bility distributions from data, Mach. Learn.: Sci. Technol. 3 (1) (2022) 015031. https://doi.org/10.1088/2632-2153/ac567a

    work page doi:10.1088/2632-2153/ac567a 2022

  47. [47]

    B. K. Natarajan, Sparse Approximate Solutions to Linear Systems, SIAM J. Comput. 24 (2) (1995) 227–234. https://doi.org/10.1137/ S0097539792240406

    1995

  48. [48]

    X. Cai, F. Nie, H. Huang, Exact top- k feature selection via ℓ2,0-norm constraint, in: Proceedings of the Twenty-Third International Joint Conference on Artificial Intelligence, Vol. 13, 2013, pp. 1240–1246

    2013

  49. [49]

    T. Pang, F. Nie, J. Han, X. Li, Efficient Feature Selection via ℓ2,0-norm Constrained Sparse Regression, IEEE Trans. Knowl. Data Eng. 31 (5) (2019) 880–893. https://doi.org/10.1109/TKDE.2018.2847685. 27

    work page doi:10.1109/tkde.2018.2847685 2019

  50. [50]

    Lemos, N

    P. Lemos, N. Jeffrey, M. Cranmer, S. Ho, P. Battaglia, Rediscovering orbital mechanics with machine learning, Mach. Learn.: Sci. Technol. 4 (4) (2022) 045021. https://doi.org/10.1088/2632-2153/acfa63

    work page doi:10.1088/2632-2153/acfa63 2022

  51. [51]

    Paszke, S

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, A. Desmaison, A. Kopf, E. Yang, Z. DeVito, M. Raison, A. Tejani, S. Chilamkurthy, B. Steiner, L. Fang, J. Bai, S. Chintala, PyTorch: An Imperative Style, High- Performance Deep Learning Library, in: Advances in Neural Information Processing Syste...

    2019

  52. [52]

    G. H. Golub, C. F. Van Loan, Matrix Computations, 4th Edition, Johns Hopkins University Press, Baltimore, MD, 2013

    2013

  53. [53]

    D. J. Korteweg, G. de Vries, XLI. On the change of form of long waves advancing in a rectangular canal, and on a new type of long stationary waves, Phil. Mag. 39 (240) (1895) 422–443. https://doi.org/10.1080/ 14786449508620739

  54. [54]

    H. Xu, D. Zhang, Robust discovery of partial differential equations in complex situations, Phys. Rev. Res. 3 (3) (2021) 033270. https://doi. org/10.1103/PhysRevResearch.3.033270

    work page doi:10.1103/physrevresearch.3.033270 2021

  55. [55]

    Stephany, C

    R. Stephany, C. Earls, Weak-PDE-LEARN: A weak form-based ap- proach to discovering PDEs from noisy, limited data, J. Comput. Phys. 506 (2024) 112950. https://doi.org/10.1016/j.jcp.2024.112950

    work page doi:10.1016/j.jcp.2024.112950 2024

  56. [56]

    İ. Dağ, Y. Dereli, Numerical solutions of KdV equation using radial basis functions, Appl. Math. Modell. 32 (4) (2008) 535–546. https: //doi.org/10.1016/j.apm.2007.02.001

    work page doi:10.1016/j.apm.2007.02.001 2008

  57. [57]

    Kassam, L

    A.-K. Kassam, L. N. Trefethen, Fourth-Order Time-Stepping for Stiff PDEs, SIAM J. Sci. Comput. 26 (4) (2005) 1214–1233. https://doi. org/10.1137/S1064827502410633

    work page doi:10.1137/s1064827502410633 2005

  58. [58]

    M. D. McKay, R. J. Beckman, W. J. Conover, Comparison of Three Methods for Selecting Values of Input Variables in the Analysis of Output from a Computer Code, Technometrics 21 (2) (1979) 239–245. https://doi.org/10.1080/00401706.1979.10489755. 28

    work page doi:10.1080/00401706.1979.10489755 1979

  59. [59]

    Tadmor, Burgers’ equation with vanishing hyper-viscosity, Com- mun

    E. Tadmor, Burgers’ equation with vanishing hyper-viscosity, Com- mun. Math. Sci. 2 (2) (2004) 317–324. https://doi.org/10.4310/ CMS.2004.V2.N2.A9

    2004

  60. [60]

    Frisch, S

    U. Frisch, S. Kurien, R. Pandit, W. Pauls, S. S. Ray, A. Wirth, J.-Z. Zhu, Hyperviscosity, galerkin truncation, and bottlenecks in turbulence, Phys. Rev. Lett. 101 (4) (2008) 144501. https://doi.org/10.1103/ PhysRevLett.101.144501

    2008

  61. [61]

    J. C. Butcher, Implicit Runge-Kutta Processes, Math. Comp. 18 (85) (1964) 50–64. https://doi.org/10.1090/ S0025-5718-1964-0159424-9

    1964

  62. [62]

    A. M. Turing, The chemical basis of morphogenesis, Philos. Trans. R. Soc. Lond. B Biol. Sci. 237 (641) (1952) 37–72. https://doi.org/10. 1098/rstb.1952.0012

    work page arXiv 1952

  63. [63]

    J. D. Murray, Mathematical Biology II: Spatial Models and Biomedical Applications, 3rd Edition, Springer, New York, 2003

    2003

  64. [64]

    J. R. Dormand, P. J. Prince, A Family of Embedded Runge-Kutta For- mulae, J. Comput. Appl. Math. 6 (1) (1980) 19–26. https://doi.org/ 10.1016/0771-050X(80)90013-3

    work page doi:10.1016/0771-050x(80)90013-3 1980

  65. [65]

    Virtanen, R

    P. Virtanen, R. Gommers, T. E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau, E. Burovski, P. Peterson, W. Weckesser, J. Bright, S. J. van der Walt, M. Brett, J. Wilson, K. J. Millman, N. Mayorov, A. R. J. Nelson, E. Jones, R. Kern, E. Larson, C. J. Carey, İ. Polat, Y. Feng, E. W. Moore, J. VanderPlas, D. Laxalde, J. Perktold, R. Cim- rman, I. Henriksen...

    2020

  66. [66]

    L. Zhao, Z. Li, B. Caswell, J. Ouyang, G. E. Karniadakis, Active learn- ing of constitutive relation from mesoscopic dynamics for macroscopic modeling of non-Newtonian flows, J. Comput. Phys. 363 (2018) 116–127. https://doi.org/10.1016/j.jcp.2018.02.039. 29

    work page doi:10.1016/j.jcp.2018.02.039 2018

  67. [67]

    Zanna and T

    L. Zanna, T. Bolton, Data-Driven Equation Discovery of Ocean Mesoscale Closures, Geophys. Res. Lett. 47 (17) (2020) e2020GL088376. https://doi.org/10.1029/2020GL088376

    work page doi:10.1029/2020gl088376 2020

  68. [68]

    Sanderse, P

    B. Sanderse, P. Stinis, R. Maulik, S. E. Ahmed, Scientific machine learn- ing for closure models in multiscale problems: A review, Found. Data Sci. 7 (1) (2025) 298–337. https://doi.org/10.3934/fods.2024043

    work page doi:10.3934/fods.2024043 2025

  69. [69]

    B. Choi, M. Ugliotti, M. Reynoso, D. R. Gurevich, R. O. Grigoriev, Data-driven modeling of multiscale phenomena with applications to fluid turbulence, Phys. Rev. E (2026). https://doi.org/10.1103/ rdj9-cjm9

    2026

  70. [70]

    Jakhar, Y

    K. Jakhar, Y. Guan, P. Hassanzadeh, Analytical and AI-Discovered Stable, Accurate, and Generalizable Subgrid-Scale Closure for Geo- physical Turbulence, Phys. Rev. Lett. 136 (2026) 064201. https: //doi.org/10.1103/v28b-5qmp. 30

    work page doi:10.1103/v28b-5qmp 2026