Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

Non-smooth optimization meets automated material model discovery

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper argues that non-smooth L1-regularized material model discovery should be solved with algorithms matched to whether the model library is linear or nonlinear in the parameters, and that this split makes computing the full…

desk verdict A solid tutorial and benchmark for L1-regularized material model discovery, with one honest but oversold new algorithm; the pathwise ISTA claim needs a caveat before this is citable as a path method. read the letter →

arxiv 2507.10196 v2 pith:7ZWDML4T submitted 2025-07-14 cs.CE cond-mat.mtrl-sci

Moritz Flaschel , Trevor Hastie , Ellen Kuhl This is my paper · ORCID
classification cs.CEcond-mat.mtrl-sci
keywords non-smoothoptimizationL1-normregularizationLASSOLARSISTAautomatedmaterialmodeldiscoveryhyperelasticitysparseregression
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

Automated material model discovery turns the search for concise constitutive laws into a sparse regression problem of the form $f(w) + \alpha\|w\|_1$, where $f$ measures the mismatch between model and data and $\alpha$ controls sparsity. The paper's central claim is that the right solver should be chosen by the structure of the model library: coordinate descent and LARS-LASSO handle libraries that are linear in the parameters, while ISTA and a pathwise extension handle nonlinear libraries. On synthetic data for incompressible hyperelastic models under uniaxial tension and simple shear, the algorithms recover known models or close surrogates and make the choice of $\alpha$ a matter of inspecting the regularization path. If this program holds, manual calibration cycles can be replaced by a principled sparse-regression step that is fast for linear libraries and reliable, given a suitable step size, for nonlinear ones.

What carries the argument

The load-bearing objects are the soft-thresholding operator $\mathrm{soft}_{\alpha}(x) = \operatorname{sign}(x)\max\{|x|-\alpha, 0\}$, which is the closed-form solution of each one-dimensional coordinate subproblem and of the proximal step in ISTA; the active-set and equiangular-direction construction of LARS-LASSO, which produces the critical values of $\alpha$ where the discovered model changes; and warm-starting along a decreasing sequence of $\alpha$ values, starting from $\alpha^{(0)} = \max_i |\partial f/\partial w_i(0)|$, in the pathwise ISTA. Together these turn the non-smooth objective $f(w)+\alpha\|w\|_1$ into a sequence of cheap, provably convergent steps.

What would settle it

Run pathwise ISTA twice on the same nonlinear hyperelastic library, once with $\alpha$ decreasing from $\alpha^{(0)}$ and once with $\alpha$ increasing from zero; if the two paths reach different sparse models at the same $\alpha$, the regularization path is initialization-dependent and would not be a reliable record of all sparse models of interest.

Watch

Extended reading notes

Core claim

The paper establishes that the four mathematical tasks of L1-regularized material model discovery require four different solvers. For libraries linear in the parameters it uses coordinate descent (CD) for a fixed regularization value, and LARS-LASSO to obtain the critical values of $\alpha$ at which material parameters enter or leave the model. For libraries nonlinear in the parameters it applies the proximal gradient method ISTA for a fixed $\alpha$, and proposes a pathwise ISTA that warm-starts each solve from the previous one while $\alpha$ decreases. The benchmark study on incompressible hyperelastic models from uniaxial tension and simple shear data shows that these tools recover Neo-Hookean, Mooney-Rivlin, Yeoh, and Ogden-type models, or sparse surrogates with low model-data mismatch, and that LARS-LASSO's early steps identify the practically relevant sparse models.

Load-bearing premise

Pathwise ISTA assumes that the solution at one regularization value is a good enough starting point for the next, so warm-starting guides a possibly non-convex objective toward the right sparse model; the paper itself notes that with different initial guesses the regularization path may not contain the models discovered by standalone ISTA.

Editorial extensions

If this is right

  • For linear material model libraries, practitioners no longer need to guess $\alpha$: LARS-LASSO identifies the critical values where the discovered model changes, so the first steps give the practically relevant sparse models.
  • For nonlinear libraries, the pathwise ISTA makes the regularization path computable by warm-starting, and the paper demonstrates this on hyperelastic benchmarks with uniaxial tension and simple shear data.
  • For mechanics applications that target models with only a few nonzero parameters, LARS-LASSO is an efficient alternative to coordinate descent, especially when it is stopped after the first few iterations.
  • A postprocessing step that re-estimates the active parameters without regularization improves fitting accuracy while leaving the discovered model structure unchanged.

Reading between the lines

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

  • The warm-start idea behind pathwise ISTA could be transferred to second-order proximal methods for nonlinear libraries, which the paper flags as a future direction rather than a demonstrated result.
  • Because the paper notes that different initial guesses can produce different regularization paths, practitioners should validate nonlinear discoveries by comparing pathwise ISTA against fixed-$\alpha$ ISTA started from several dense initial guesses.
  • The critical values produced by LARS-LASSO suggest a principled alternative to trial-and-error or Pareto-based selection of $\alpha$ in material discovery, though the paper does not itself make that model-selection claim.
  • When features in a linear library are highly correlated, the matrix $\bar{X}_A^T \bar{X}_A$ inverted in LARS-LASSO can become ill-conditioned; the paper's early stopping mitigates this, and alternative active-set updates would be a natural robustness extension.
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

3 major / 4 minor

Summary. The paper studies L1-regularized sparse regression for automated material model discovery. It formulates four problems: the LASSO for a fixed regularization parameter (Problem 1), the computation of the LASSO critical values and regularization path (Problem 2), the L1-regularized nonlinear model-data mismatch (Problem 3), and the computation of a regularization path for the nonlinear case (Problem 4). It then reviews coordinate descent and LARS/LARS-LASSO for the linear case, and ISTA and a warm-started pathwise ISTA for the nonlinear case. Benchmarks on synthetic uniaxial-tension and simple-shear data for incompressible hyperelastic models (Neo-Hookean, Mooney-Rivlin, Yeoh, Biderman, Ogden, and a mixed model) show successful recovery in several cases and honest failures in others. Code and data are provided on Zenodo and GitHub.

Significance. The paper's main value is as a careful, well-documented connection between established non-smooth optimization methods and the material-model-discovery literature, with reproducible implementations and honest reporting of recovery failures. The derivations for coordinate descent, LARS, and ISTA are standard and appear correct, and the benchmark tables are informative. The genuinely new component, pathwise ISTA for nonconvex objectives, is not yet convincing: the claim that it computes the regularization path is undermined by the paper's own observation that the path depends on the initialization. With a revised, more circumscribed statement of what Algorithm 4 provides, the paper would be a useful reference for practitioners.

major comments (3)
  1. [§3.4, §5.2, §6] The central claim in §3.4 and §6 that Algorithm 4 'efficiently computes the nonlinear regularization path' is not supported for nonconvex f. For such f, the minimizer set of Eq. (10) is not single-valued, and ISTA is only guaranteed to converge to a stationary point of the composite objective from a given initialization (Beck, 2017). Algorithm 4 therefore computes a single warm-start trajectory, not a well-defined path. Section 5.2 explicitly states that 'due to the different initial guesses, the regularization path may not contain the models discovered by ISTA', which concedes the point. This is not a mere caveat: a practitioner following the pathwise recommendation can miss the exact sparse model that single-alpha ISTA with another equally valid initialization produces. Please reframe Algorithm 4 as a heuristic continuation method, state conditions under which warm starts track the true path, or provide a verification procedure and benchmarks showing that the selected models are robust to initialization.
  2. [§5.2, Table 5 vs. Figure 6] The initialization dependence is concretely visible in the benchmarks. For the Mooney-Rivlin dataset, Table 5 (ISTA with w^(0)=1) discovers a model with an Ogden feature, while the text reports that pathwise ISTA 'correctly identifies the Mooney-Rivlin features' for the same benchmark; for the Ogden and mixed benchmarks, pathwise ISTA misses the Ogden feature. The paper notes this discrepancy but does not quantify it or tell the reader which output should be trusted for model selection. The authors should report the actual models on the pathwise solution at selected alpha values, compare pathwise and single-alpha ISTA solutions at the same alpha, and explain the practical selection rule that resolves this ambiguity.
  3. [§3.4, Algorithm 4] The claimed computational advantage of warm starting is not demonstrated. No runtimes, iteration counts, or comparisons with n_alpha independent ISTA solves are reported, and the step size gamma and grid size n_alpha are free parameters whose influence on the computed path is not discussed. Since the conclusion in §6 rests on the efficiency of Algorithm 4, the authors should add at least a sensitivity analysis for gamma and n_alpha and an iteration-count or runtime comparison against independent ISTA solves.
minor comments (4)
  1. [Algorithm 1] The convergence check in Algorithm 1 refers to w^(k+1) before it is defined; the pseudocode should assign the result of the coordinate sweep to w^(k+1) before testing the stopping criterion.
  2. [§4.3, Eq. (54)] The norm in the definition of f(w) should be written as ||y - Xw||_2^2 for consistency with Eq. (6) and with Eq. (51).
  3. [§3.4] The sentence 'the initial guess w^(l)(k) = w^(l-1)' should read 'the initial guess w^(l)(0) = w^(l-1)' to match Algorithm 4.
  4. [§5.2, Figure 6] The text around Figure 6 would benefit from a table listing the actual pathwise models at selected alpha values, so that the reader can directly see which features enter and leave the path and how they compare with the Table 5 ISTA results.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: CD, LARS-LASSO, and ISTA are standard algorithms validated against synthetic ground-truth models; the pathwise-ISTA limitation noted in Section 5.2 is a robustness concern, not a circular derivation.

full rationale

The paper's derivations are self-contained and do not reduce to their inputs. Coordinate descent, LARS-LASSO, and ISTA are standard methods cited to independent foundational sources (Tibshirani 1996; Efron et al. 2004; Beck 2017), and the paper's contribution is to apply them to material-model-discovery libraries. The benchmarks are generated from known ground-truth material models, and the algorithms are assessed by whether they recover those models; no fitted parameter is later relabeled as a prediction. Self-citations to Flaschel et al. appear in the framing and library construction, but the load-bearing algorithmic content does not depend on them. The one notable flagged limitation is in Section 5.2: 'due to the different initial guesses, the regularization path may not contain the models discovered by ISTA.' This correctly concedes that Algorithm 4 computes an initialization-dependent warm-start trajectory rather than a uniquely defined nonlinear regularization path. That is a limitation on the efficiency claim in Sections 3.4 and 6, but it is not circular: the algorithm does exactly what it states, and the concession does not redefine any output as an input. Accordingly, no circularity pattern is exhibited, and the score is 0.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claims rest on standard optimization assumptions (differentiability, Lipschitz gradients, linear independence), plus two ad hoc heuristics: the existence of ordered critical values and the warm-starting logic in pathwise ISTA. No new physical entities are introduced. The main hidden choices are the ISTA step size and the grid size n_alpha, both of which affect the reported results.

free parameters (2)
  • ISTA step size gamma
    Convergence of ISTA depends on choosing a step size bounded by the inverse Lipschitz constant of grad f (Section 3.3). The paper specifies no concrete value or rule in the text and refers to trial and error, so a practitioner cannot reproduce the shown paths without inspecting the code.
  • Pathwise grid size n_alpha = 1000
    The pathwise ISTA experiments (Section 5.2) use n_alpha = 1000, a hand-chosen hyperparameter that controls the resolution of the computed regularization path.
assumptions (4)
  • domain assumption Feature vectors X_i are linearly independent.
    Assumed in Section 2.1 and used for the OLS initialization of CD (Algorithm 1), for the projection y_parallel = X[X^T X]^{-1} X^T y in LARS, and for the equiangular-vector inversion in Eq. (25).
  • ad hoc to paper For each critical sparsity level c, there exists an ordering of critical alpha_c with simultaneous zero-crossings neglected.
    Stated in the footnote to Problem 2: 'For simplicity, we assume the existence of the sequence of values alpha_c' and 'we thus neglect this special case.'
  • domain assumption f(w) is differentiable and grad f is Lipschitz continuous for Problems 3 and 4.
    Assumed in Section 2.3 and used for the ISTA convergence guarantee quoted from Beck (2017) and for the gradient computation via automatic differentiation.
  • ad hoc to paper Warm starts from the previous alpha-solution provide good initial guesses along the path.
    Core heuristic behind Pathwise ISTA (Algorithm 4), inspired by Friedman et al. (2007, 2010). The paper flags in Section 5.2 that the path may miss local minima found from other initial guesses.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Non-smooth optimization meets automated material model discovery." pith.science (2026). https://pith.science/paper/7ZWDML4T

@misc{pith2026250710196,
  author       = {Pith},
  title        = {Pith review of: Non-smooth optimization meets automated material model discovery},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7ZWDML4T}},
  note         = {Machine review of arXiv:2507.10196}
}
read the original abstract

Automated material model discovery disrupts the tedious and time-consuming cycle of iteratively calibrating and modifying manually designed models. Non-smooth L1-norm regularization is the backbone of automated model discovery; however, the current literature on automated material model discovery offers limited insights into the robust and efficient minimization of non-smooth objective functions. In this work, we examine the minimization of functions of the form f(w) + a ||w||_1, where w are the material model parameters, f is a metric that quantifies the mismatch between the material model and the observed data, and a is a regularization parameter that determines the sparsity of the solution. We investigate both the straightforward case where f is quadratic and the more complex scenario where it is non-quadratic or even non-convex. Importantly, we do not only focus on methods that solve the sparse regression problem for a given value of the regularization parameter a, but propose methods to efficiently compute the entire regularization path, facilitating the selection of a suitable a. Specifically, we present four algorithms and discuss their roles for automated material model discovery in mechanics: First, we recapitulate a well-known coordinate descent algorithm that solves the minimization problem assuming that f is quadratic for a given value of a, also known as the LASSO. Second, we discuss the algorithm LARS, which automatically determines the critical values of a, at which material parameters in w are set to zero. Third, we propose to use the proximal gradient method ISTA for automated material model discovery if f is not quadratic, and fourth, we suggest a pathwise extension of ISTA for computing the regularization path. We demonstrate the applicability of all algorithms for the discovery of hyperelastic material models from uniaxial tension and simple shear data.

Figures

Figures reproduced from arXiv: 2507.10196 by the authors.

Figure 1
Figure 1. Qualitative regularization path of Problem [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the first step of LARS considering two features. All vectors are illustrated in the two-dimensional plane spanned by the [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Regularization path computed by LARS-LASSO for the noise-free Yeoh dataset. For clarity, legend entries of higher order features are [PITH_FULL_IMAGE:figures/full_fig_p022_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Regularization path computed by LARS-LASSO for the noisy Yeoh dataset. For clarity, legend entries of higher order features are [PITH_FULL_IMAGE:figures/full_fig_p023_4.png]
Figure 5
Figure 5. Figure 5: Stress-strain response of the discovered material model for the noisy Yeoh dataset. [PITH_FULL_IMAGE:figures/full_fig_p024_5.png]
Figure 6
Figure 6. Figure 6: Regularization path computed by the pathwise ISTA for the noisy datasets. For clarity, legend entries of higher order Mooney-Rivlin [PITH_FULL_IMAGE:figures/full_fig_p025_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Adaptive Material Fingerprinting for the fast discovery of polyconvex feature combinations in isotropic and anisotropic hyperelasticity

    cs.CE 2026-04 unverdicted novelty 7.0 of 10

    An adaptive database and iterative pattern recognition algorithm lets Material Fingerprinting discover arbitrary linear combinations of polyconvex isotropic and anisotropic hyperelastic features from experimental data.

Reference graph

Works this paper leans on

61 extracted references · 22 canonical work pages · cited by 1 Pith paper

  1. [1]

    Friedman, J., Hastie, T., Tibshirani, R.,

    URL: https:// projecteuclid.org/journals/annals-of-applied-statistics/volume-1/issue-2/Pathwise-coordinate-optimization/ 10.1214/07-AOAS131.full, doi:10.1214/07-AOAS131. Friedman, J., Hastie, T., Tibshirani, R.,

    work page doi:10.1214/07-aoas131.full

  2. [4]

    Linka, K., Hillg ¨artner, M., Abdolazizi, K.P., Aydin, R.C., Itskov, M., Cyron, C.J.,

    URL: http: //link.springer.com/10.1007/BF00175355, doi:10.1007/BF00175355. Linka, K., Hillg ¨artner, M., Abdolazizi, K.P., Aydin, R.C., Itskov, M., Cyron, C.J.,

    doi:10.1007/bf00175355

  3. [5]

    Computer Methods in Applied Mechanics and Engineering 425, 116967

    Unsupervised learning of history-dependent constitutive material laws with thermodynamically- consistent neural networks in the modified Constitutive Relation Error framework. Computer Methods in Applied Mechanics and Engineering 425, 116967. URL: https://linkinghub.elsevier.com/retrieve/pii/S0045782524002238, doi:10.1016/j.cma.2024.116967. Bleyer, J., 20...

    arXiv

  4. [6]

    Bomarito, G., Townsend, T., Stewart, K., Esham, K., Emery, J., Hochhalter, J.,

    Learning elastoplasticity with implicit layers doi:http://dx.doi.org/10.2139/ssrn.5210734. Bomarito, G., Townsend, T., Stewart, K., Esham, K., Emery, J., Hochhalter, J.,

    work page doi:10.2139/ssrn.5210734

  5. [8]

    Science Advances 7, eabf3658

    One for all: Universal material model based on minimal state-space neural networks. Science Advances 7, eabf3658. URL: https://advances.sciencemag.org/lookup/doi/10.1126/sciadv.abf3658, doi:10.1126/sciadv.abf3658. Boyd, S.P., Vandenberghe, L.,

    doi:10.1126/sciadv.abf3658

  6. [12]

    Automated Discovery of Material Models in Continuum Solid Mechanics. Ph.D. thesis. ETH Zurich. URL: http://hdl. handle.net/20.500.11850/602750, doi:10.3929/ETHZ-B-000602750 . Flaschel, M., Hastie, T., Kuhl, E., 2025a. Supplementary software for ”Non-smooth optimization meets automated material model discovery”. Zenodo doi:https://doi.org/10.5281/zenodo.15...

    doi:10.3929/ethz-b-000602750

  7. [13]

    Computer Methods in Applied Mechanics and Engineering 381, 113852

    Unsupervised discovery of interpretable hyperelastic constitutive laws. Computer Methods in Applied Mechanics and Engineering 381, 113852. doi: 10.1016/j.cma.2021.113852. Flaschel, M., Kumar, S., De Lorenzis, L.,

    work page arXiv 2021

  8. [19]

    Computer Methods in Applied Mechanics and Engineering 388, 114217

    Local approximate Gaussian process regression for data-driven constitutive models: development and comparison with neural networks. Computer Methods in Applied Mechanics and Engineering 388, 114217. URL: https://linkinghub. elsevier.com/retrieve/pii/S004578252100548X, doi:10.1016/j.cma.2021.114217. Ghaboussi, J., Garrett, J.H., Wu, X.,

    arXiv

Show all 61 references
  1. [23]

    URL: http://arxiv.org/abs/2410.13645

    Automated Model Discovery for Tensional Homeostasis: Constitutive Machine Learning in Growth and Remodeling. URL: http://arxiv.org/abs/2410.13645. arXiv:2410.13645 [cs]. Holzapfel, G.A.,

    arXiv

  2. [25]

    Applications in Engineering Science 6, 100052

    Application of symbolic regression for constitutive modeling of plastic deformation. Applications in Engineering Science 6, 100052. URL: https://linkinghub.elsevier.com/retrieve/pii/ S2666496821000182, doi:10.1016/j.apples.2021.100052. Kalina, K.A., Linden, L., Brummund, J., M...

  3. [26]

    Computational Mechanics 69, 213–232

    Automated constitutive modeling of isotropic hyperelasticity based on artificial neural networks. Computational Mechanics 69, 213–232. URL: https://link.springer.com/10.1007/s00466-021-02090-6 , doi:10.1007/s00466-021-02090-6 . Kanno, Y .,

    doi

  4. [28]

    Computer Methods in Applied Mechanics and Engineering 304, 81–101

    Data-driven computational mechanics. Computer Methods in Applied Mechanics and Engineering 304, 81–101. URL: https://linkinghub.elsevier.com/retrieve/pii/S0045782516300238, doi:10.1016/j.cma.2016.02.001. Kissas, G., Mishra, S., Chatzi, E., De Lorenzis, L.,

    2016 doi

  5. [29]

    Computer Methods in Applied Mechanics and Engineering 428, 117053

    The language of hyperelastic materials. Computer Methods in Applied Mechanics and Engineering 428, 117053. URL: https://linkinghub.elsevier.com/retrieve/pii/S0045782524003098, doi:10.1016/j.cma. 2024.117053. Klein, D.K., Fern ´andez, M., Martin, R.J., Ne ff, P., Weeger, O.,

    2024

  6. [30]

    Journal of the Mechanics and Physics of Solids 159, 104703

    Polyconvex anisotropic hyperelasticity with neural networks. Journal of the Mechanics and Physics of Solids 159, 104703. URL: https://linkinghub.elsevier.com/retrieve/pii/S0022509621003215, doi:10.1016/j.jmps.2021.104703. Koza, J.,

    2021

  7. [32]

    Journal of Computational Physics 429, 110010

    Constitutive artificial neural networks: A fast and general approach to predictive data-driven constitutive modeling by deep learning. Journal of Computational Physics 429, 110010. URL: https://linkinghub.elsevier.com/retrieve/pii/S0021999120307841, doi:10.1016/j.jcp.2020.1100...

  8. [33]

    URL: http://www.jstatsoft.org/v33/i01/, doi:10.18637/jss.v033.i01. Fu, W.,

    doi

  9. [34]

    Computer Methods in Applied Mechanics and Engineering 403, 115731

    A new family of Constitutive Artificial Neural Networks towards automated model discovery. Computer Methods in Applied Mechanics and Engineering 403, 115731. URL: https://linkinghub.elsevier.com/retrieve/pii/S0045782522006867, doi:10.1016/j.cma.2022.115731. Linka, K., Kuhl, E.,

  10. [35]

    Extreme Mechan- ics Letters 70, 102181

    Best-in-class modeling: A novel strategy to discover constitutive models for soft matter systems. Extreme Mechan- ics Letters 70, 102181. URL: https://linkinghub.elsevier.com/retrieve/pii/S2352431624000610, doi:10.1016/j.eml.2024. 102181. Linka, K., St. Pierre, S.R., Kuhl, E.,

    doi

  11. [36]

    Acta Bio- materialia 160, 134–151

    Automated model discovery for human brain using Constitutive Artificial Neural Networks. Acta Bio- materialia 160, 134–151. URL: https://linkinghub.elsevier.com/retrieve/pii/S1742706123000661, doi:10.1016/j.actbio. 2023.01.055. Marckmann, G., Verron, E.,

    2023 doi

  12. [37]

    32 Mechanics of Materials 181, 104643

    Automated identification of linear viscoelastic constitutive laws with EUCLID. 32 Mechanics of Materials 181, 104643. URL: https://linkinghub.elsevier.com/retrieve/pii/S0167663623000893, doi:10.1016/ j.mechmat.2023.104643. Masi, F., Stefanou, I., Vannucci, P., Maffi-Berthier, V .,

    2023

  13. [38]

    Journal of the Mechanics and Physics of Solids 147, 104277

    Thermodynamics-based Artificial Neural Networks for constitutive modeling. Journal of the Mechanics and Physics of Solids 147, 104277. URL: https://linkinghub.elsevier.com/retrieve/pii/S0022509620304841, doi:10.1016/j.jmps.2020.104277. McCulloch, J.A., St. Pierre, S.R., Linka,...

  14. [39]

    International Journal for Numerical Methods in Engineering 125, e7481

    On sparse regression,Lp-regularization, and automated model discovery. International Journal for Numerical Methods in Engineering 125, e7481. URL:https://onlinelibrary.wiley.com/doi/10.1002/nme.7481, doi:10. 1002/nme.7481. Meyer, K.A., Ekre, F.,

    doi

  15. [40]

    URL: https: //engrxiv.org/preprint/view/2961, doi:https://doi.org/10.31224/2961

    Thermodynamically consistent neural network plasticity modeling and discovery of evolution laws. URL: https: //engrxiv.org/preprint/view/2961, doi:https://doi.org/10.31224/2961. Moon, H., Park, D., Cho, H., Noh, H.K.,

    doi

  16. [41]

    arXiv:2504.19494 URL: https://arxiv.org/abs/2504.19494

    Physics-Informed Neural Network-Based Discovery of Hyperelastic Constitutive Models from Extremely Scarce Data. arXiv:2504.19494 URL: https://arxiv.org/abs/2504.19494. Ogden, R.W.,

  17. [43]

    Composites Part B: Engineering 216, 108831

    Multiscale constitutive model using data–driven yield function. Composites Part B: Engineering 216, 108831. URL: https://linkinghub.elsevier.com/retrieve/pii/S1359836821002225, doi:10.1016/j.compositesb.2021.108831. Peirlinck, M., Linka, K., Hurtado, J.A., Holzapfel, G.A., Kuhl, E.,

    2021

  18. [44]

    Computational Mechanics URL: https://link.springer.com/10.1007/s00466-024-02515-y , doi:10.1007/s00466-024-02515-y

    Democratizing biomedical simulation through automated model discovery and a universal material subroutine. Computational Mechanics URL: https://link.springer.com/10.1007/s00466-024-02515-y , doi:10.1007/s00466-024-02515-y . Ratle, A., Sebag, M.,

    doi

  19. [45]

    Applied Soft Computing 1, 105–118

    Grammar-guided genetic programming and dimensional consistency: application to non-parametric identification in mechanics. Applied Soft Computing 1, 105–118. URL: https://linkinghub.elsevier.com/retrieve/pii/S1568494601000096, doi:10.1016/S1568-4946(01)00009-6 . Rivlin,

    doi

  20. [47]

    International Journal for Numerical Methods in Engineering 124, 4802–4840

    A comparative study on di fferent neural network architectures to model inelas- ticity. International Journal for Numerical Methods in Engineering 124, 4802–4840. URL: https://onlinelibrary.wiley.com/doi/10. 1002/nme.7319, doi:10.1002/nme.7319. Santosa, F., Symes, W.W.,

    doi

  21. [49]

    Science 324, 81–85

    Distilling Free-Form Natural Laws from Experimental Data. Science 324, 81–85. URL: https://www. sciencemag.org/lookup/doi/10.1126/science.1165893, doi:10.1126/science.1165893. Schoenauer, M., Sebag, M., Jouve, F., Lamy, B., Maitournam, H.,

    doi

  22. [50]

    Pierre, S.R., Linka, K., Kuhl, E., 2023a

    St. Pierre, S.R., Linka, K., Kuhl, E., 2023a. Principal-stretch-based constitutive neural networks autonomously discover a subclass of Ogden models for human brain tissue. Brain Multiphysics 4, 100066. URL:https://linkinghub.elsevier.com/retrieve/pii/S2666522023000047, doi:10....

  23. [51]

    Communications in Numerical Methods in Engineering 25, 53–63

    A model of incompressible isotropic hyperelastic material behavior using spline interpolations of tension- compression test data. Communications in Numerical Methods in Engineering 25, 53–63. URL: http://doi.wiley.com/10.1002/cnm. 1105, doi:10.1002/cnm.1105. Tac, V ., Sahli Co...

    doi

  24. [52]

    Com- puter Methods in Applied Mechanics and Engineering 398, 115248

    Data-driven tissue mechanics with polyconvex neural ordinary di fferential equations. Com- puter Methods in Applied Mechanics and Engineering 398, 115248. URL: https://linkinghub.elsevier.com/retrieve/pii/ S0045782522003838, doi:10.1016/j.cma.2022.115248. Thakolkaran, P., Guo,...

  25. [53]

    URL:http://arxiv.org/abs/2503.05617, doi:10.48550/arXiv

    Can KAN CANs? Input-convex Kolmogorov-Arnold Networks (KANs) as hyperelastic constitutive artificial neural networks (CANs). URL:http://arxiv.org/abs/2503.05617, doi:10.48550/arXiv. 2503.05617. arXiv:2503.05617 [cs]. Thakolkaran, P., Joshi, A., Zheng, Y ., Flaschel, M., De Lor...

    doi arXiv

  26. [54]

    Journal of the Mechanics and Physics of Solids 169, 105076

    NN-EUCLID: Deep-learning hyperelasticity without stress data. Journal of the Mechanics and Physics of Solids 169, 105076. URL: https://linkinghub.elsevier.com/retrieve/pii/ S0022509622002538, doi:10.1016/j.jmps.2022.105076. Tibshirani, R.,

    2022

  27. [57]

    Computer Methods in Applied Mechanics and Engineering 318, 981–1004

    Data driven modeling of plastic deformation. Computer Methods in Applied Mechanics and Engineering 318, 981–1004. URL: https://linkinghub.elsevier.com/retrieve/pii/S0045782516314499, doi:10.1016/j.cma. 2017.02.016. Vlassis, N., Ma, R., Sun, W.,

    2017 doi

  28. [58]

    arXiv:2001.04292 [cs] URL: http://arxiv.org/abs/2001.04292, doi:10.1016/j.cma.2020.113299

    Geometric deep learning for computational mechanics Part I: Anisotropic Hyperelasticity. arXiv:2001.04292 [cs] URL: http://arxiv.org/abs/2001.04292, doi:10.1016/j.cma.2020.113299. arXiv: 2001.04292. Wang, M., Chen, C., Liu, W.,

    2001 arXiv

  29. [59]

    Journal of the Mechanics and Physics of Solids 153, 104474

    Inference of deformation mechanisms and constitutive response of soft material surrogates of biological tissue by full-field characterization and data-driven variational system identification. Journal of the Mechanics and Physics of Solids 153, 104474. URL: https://linkinghub....

  30. [60]

    Yang, J., Hastie, T., 2024a

    URL: https://amses-journal.springeropen.com/articles/10.1186/ s40323-024-00281-3 , doi:10.1186/s40323-024-00281-3 . Yang, J., Hastie, T., 2024a. A Fast and Scalable Pathwise-Solver for Group Lasso and Elastic Net Penalized Regression via Block-Coordinate Descent. URL: http://a...

    doi arXiv

  31. [91]

    Flaschel, M., Kumar, S., De Lorenzis, L., 2023a

    URL: https://www.nature.com/articles/s41524-022-00752-4 , doi:10.1038/s41524-022-00752-4 . Flaschel, M., Kumar, S., De Lorenzis, L., 2023a. Automated discovery of generalized standard material models with EUCLID. Computer Methods in Applied Mechanics and Engineering 405, 11586...

  32. [1951]

    Experiments on the deformation of rubber

    Large elastic deformations of isotropic materials VII. Experiments on the deformation of rubber. Philosophical Transactions of the Royal Society of London. Series A, Mathematical and Physical Sciences 243, 251–288. URL: https://royalsocietypublishing.org/ doi/10.1098/rsta.1951...

  33. [1972]

    Proceedings of the Royal Society of London

    Large deformation isotropic elasticity – on the correlation of theory and experiment for incompressible rubberlike solids. Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences, , 565–584. Osborne, M., Presnell, B., Turlach, B.A., 2000a. A ne...

    doi

  34. [1986]

    SIAM Journal on Scientific and Statistical Computing 7, 1307–1330

    Linear Inversion of Band-Limited Reflection Seismograms. SIAM Journal on Scientific and Statistical Computing 7, 1307–1330. URL: http://epubs.siam.org/doi/10.1137/0907087, doi:10.1137/0907087. Schmidt, M., Lipson, H.,

    doi

  35. [1991]

    Journal of Engineer- ing Mechanics 117, 132–153

    Knowledge-Based Modeling of Material Behavior with Neural Networks. Journal of Engineer- ing Mechanics 117, 132–153. URL:http://ascelibrary.org/doi/10.1061/%28ASCE%290733-9399%281991%29117%3A1%28132%29, doi:10.1061/(ASCE)0733-9399(1991)117:1(132). Hartmann, S.,

    1991 doi

  36. [1993]

    Technometrics 35, 109–135

    A Statistical View of Some Chemometrics Regression Tools. Technometrics 35, 109–135. URL: http: //www.tandfonline.com/doi/abs/10.1080/00401706.1993.10485033, doi:10.1080/00401706.1993.10485033. Friedman, J., Hastie, T., H¨ofling, H., Tibshirani, R.,

    1993

  37. [1996]

    Journal of the Royal Statistical Society: Series B (Methodological) 58, 267–288

    Regression Shrinkage and Selection via the Lasso. Journal of the Royal Statistical Society: Series B (Methodological) 58, 267–288. URL: http://doi.wiley.com/10.1111/j.2517-6161.1996.tb02080.x, doi:10.1111/j.2517-6161.1996.tb02080.x. Udrescu, S.M., Tegmark, M.,

    1996

  38. [1998]

    Fuhg, J.N., Anantha Padmanabha, G., Bouklas, N., Bahmani, B., Sun, W., Vlassis, N.N., Flaschel, M., Carrara, P., De Lorenzis, L., 2024a

    Penalized Regressions: The Bridge versus the Lasso . Fuhg, J.N., Anantha Padmanabha, G., Bouklas, N., Bahmani, B., Sun, W., Vlassis, N.N., Flaschel, M., Carrara, P., De Lorenzis, L., 2024a. A Review on Data-Driven Constitutive Laws for Solids. Archives of Computational Methods...

  39. [2001]

    International Journal of Solids and Structures 38, 7999–8018

    Parameter estimation of hyperelasticity relations of generalized polynomial-type with constraint conditions. International Journal of Solids and Structures 38, 7999–8018. URL: https://linkinghub.elsevier.com/retrieve/pii/S002076830100018X, doi:10.1016/S0020-7683(01)00018-X . H...

    doi

  40. [2004]

    Com- munications on Pure and Applied Mathematics 57, 1413–1457

    An iterative thresholding algorithm for linear inverse problems with a sparsity constraint. Com- munications on Pure and Applied Mathematics 57, 1413–1457. URL: https://onlinelibrary.wiley.com/doi/10.1002/cpa.20042, doi:10.1002/cpa.20042. Dubˇc´akov´a, R.,

    doi

  41. [2005]

    Journal of the Royal Statistical Society: Series B (Statisti- cal Methodology) 67, 301–320

    Regularization and variable selection via the elastic net. Journal of the Royal Statistical Society: Series B (Statisti- cal Methodology) 67, 301–320. URL: http://doi.wiley.com/10.1111/j.1467-9868.2005.00503.x, doi: 10.1111/j.1467-9868. 2005.00503.x. 34

    2005

  42. [2006]

    Rubber Chem- istry and Technology 79, 835–858

    Comparison of Hyperelastic Models for Rubber-Like Materials. Rubber Chem- istry and Technology 79, 835–858. URL: https://meridian.allenpress.com/rct/article/79/5/835/93139/ Comparison-of-Hyperelastic-Models-for-RubberLike , doi:10.5254/1.3547969. Marino, E., Flaschel, M., Kuma...

    doi

  43. [2007]

    IEEE Journal of Selected Topics in Signal Processing 1, 606–617

    An Interior-Point Method for Large-Scale $ \ell 1$-Regularized Least Squares. IEEE Journal of Selected Topics in Signal Processing 1, 606–617. URL: http://ieeexplore.ieee.org/document/4407767/, doi:10. 1109/JSTSP.2007.910971. Kirchdoerfer, T., Ortiz, M.,

    2007

  44. [2009]

    Springer Series in Statistics, Springer New York, New York, NY

    The Elements of Statistical Learning: Data Mining, Inference, and Prediction. Springer Series in Statistics, Springer New York, New York, NY . URL: http://link.springer.com/10.1007/978-0-387-84858-7 , doi: 10.1007/ 978-0-387-84858-7 . Holthusen, H., Brepols, T., Linka, K., Kuhl, E.,

    doi

  45. [2015]

    Journal of Elasticity 120, 129–160

    Hyperelastic Energy Densities for Soft Biological Tissues: A Review. Journal of Elasticity 120, 129–160. URL: http://link.springer.com/10.1007/s10659-014-9508-z , doi:10.1007/s10659-014-9508-z . Dal, H., Ac ¸ıkg ¨oz, K., Badienia, Y .,

    doi

  46. [2016]

    Proceedings of the National Academy of Sciences 113, 3932–3937

    Discovering governing equations from data by sparse identification of nonlinear dynamical sys- tems. Proceedings of the National Academy of Sciences 113, 3932–3937. URL: http://www.pnas.org/lookup/doi/10.1073/pnas. 1517384113, doi:10.1073/pnas.1517384113. Chagnon, G., Rebouah,...

    doi

  47. [2017]

    Computational Mechanics 60, 813–826

    Data-driven non-linear elasticity: constitutive manifold construction and problem discretization. Computational Mechanics 60, 813–826. URL: http://link.springer.com/ 10.1007/s00466-017-1440-1 , doi:10.1007/s00466-017-1440-1 . James, G., Witten, D., Hastie, T., Tibshirani, R., ...

    doi

  48. [2020]

    Science Advances 6, eaay2631

    AI Feynman: A physics-inspired method for symbolic regression. Science Advances 6, eaay2631. URL: 33 https://advances.sciencemag.org/lookup/doi/10.1126/sciadv.aay2631, doi:10.1126/sciadv.aay2631. Versino, D., Tonda, A., Bronkhorst, C.A.,

    doi

  49. [2021]

    Computers & Structures 252, 106557

    Development of interpretable, data-driven plasticity models with symbolic regression. Computers & Structures 252, 106557. URL: https://linkinghub.elsevier.com/retrieve/pii/ S0045794921000791, doi:10.1016/j.compstruc.2021.106557. Bonatti, C., Mohr, D.,

    2021

  50. [2022]

    net/10.13140/RG.2.2.21694.36168, doi:10.13140/RG.2.2.21694.36168

    A Mechanics-Informed Neural Network Framework for Data-Driven Nonlinear Viscoelasticity URL:https://rgdoi. net/10.13140/RG.2.2.21694.36168, doi:10.13140/RG.2.2.21694.36168. publisher: Unpublished. Bahmani, B., Sun, W.,

  51. [2023]

    International Journal for Numerical Methods in Engineering , nme.7203URL: https://onlinelibrary.wiley.com/doi/10.1002/nme

    Automatic generation of interpretable hyperelastic material models by symbolic regression. International Journal for Numerical Methods in Engineering , nme.7203URL: https://onlinelibrary.wiley.com/doi/10.1002/nme. 7203, doi:10.1002/nme.7203. As’ad, F., Farhat, C.,

    doi

  52. [2024]

    International Journal for Numerical Methods in Engineering 125, e7473

    Physics-constrained symbolic model discovery for polyconvex incompressible hyperelastic materials. International Journal for Numerical Methods in Engineering 125, e7473. URL:https://onlinelibrary.wiley.com/doi/10.1002/nme.7473, doi:10. 1002/nme.7473. Beck, A.,

    doi

  53. [2025]

    URL: http://arxiv.org/abs/2502.05682, doi:10.48550/arXiv.2502.05682

    Constitutive Kolmogorov-Arnold Networks (CKANs): Combining Accuracy and Interpretability in Data-Driven Material Modeling. URL: http://arxiv.org/abs/2502.05682, doi:10.48550/arXiv.2502.05682. arXiv:2502.05682 [physics]. Abdusalamov, R., Hillg¨artner, M., Itskov, M.,

    doi arXiv

Pith tools

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