pith. machine review for the scientific record. sign in

arxiv: 2603.27351 · v2 · submitted 2026-03-28 · 💻 cs.CE

Recognition: 2 theorem links

· Lean Theorem

Modeling isotropic polyconvex hyperelasticity by neural networks -- sufficient and necessary criteria for compressible and incompressible materials

Gian-Luca Geuken , Patrick Kurzeja , David Wiedemann , Martin Zlati\'c , Marko \v{C}ana{\dj}ija , J\"orn Mosler
Authors on Pith no claims yet

Pith reviewed 2026-05-14 21:40 UTC · model grok-4.3

classification 💻 cs.CE
keywords hyperelasticitypolyconvexityneural networksisotropic materialscompressible materialsincompressible materialsfinite strainstrain energy function
0
0 comments X

The pith

Convex signed singular value neural networks provide universal approximation for frame-indifferent isotropic polyconvex hyperelastic energies in both compressible and incompressible regimes.

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

The paper establishes that CSSV-NNs and their incompressible variant inc-CSSV-NNs can represent any frame-indifferent isotropic polyconvex strain-energy function to arbitrary accuracy. This is achieved by enforcing convexity directly on the signed singular values of the deformation gradient, avoiding the overly narrow constraints used in earlier neural-network formulations. The approach is tested on classical hyperelastic models including Neo-Hooke, Mooney-Rivlin, Gent and Arruda-Boyce, as well as Treloar’s rubber data, and is shown to recover these targets accurately. An explicit counterexample demonstrates that Ball’s sufficient criterion for polyconvexity excludes some valid energies that the new networks can still represent.

Core claim

CSSV-NNs achieve universal approximation of frame-indifferent isotropic polyconvex energies for compressible materials by placing convexity constraints on the signed singular values of the deformation gradient; the inc-CSSV-NN version extends the same guarantee to the incompressible case by incorporating the determinant constraint directly into the network architecture. In contrast, earlier neural-network models that rely on Ball’s criterion or similar sufficient conditions impose restrictions that exclude admissible polyconvex functions, thereby limiting expressiveness. Numerical evidence confirms that the proposed networks recover established analytical models and experimental data while a

What carries the argument

Convex Signed Singular Value Neural Networks (CSSV-NNs) that enforce polyconvexity through convexity constraints applied to the signed singular values of the deformation gradient.

If this is right

  • Any classical or data-driven isotropic polyconvex energy can be represented without manual derivation of a closed-form expression.
  • The same network class works for both compressible and incompressible materials without additional ad-hoc corrections.
  • Fitting to experimental stress-strain data becomes possible while automatically satisfying frame-indifference, isotropy and polyconvexity.
  • Ball’s criterion is shown to be sufficient but not necessary, allowing a strictly larger set of admissible models.
  • Finite-element implementations can directly use the trained networks as constitutive routines.

Where Pith is reading between the lines

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

  • The method could be combined with automatic differentiation to obtain consistent stress and tangent operators for large-scale simulations.
  • Extension to anisotropic materials would require adding additional network inputs that respect the material symmetry group.
  • The explicit counterexample to Ball’s criterion suggests that new, weaker necessary conditions for polyconvexity may be discoverable by examining the network’s functional form.

Load-bearing premise

The chosen neural-network architecture with signed singular values and convexity constraints can be trained to approximate any target polyconvex function to arbitrary accuracy.

What would settle it

A specific isotropic frame-indifferent polyconvex energy function that a CSSV-NN of finite width and depth cannot approximate to within a small error tolerance after exhaustive training.

Figures

Figures reproduced from arXiv: 2603.27351 by David Wiedemann, Gian-Luca Geuken, J\"orn Mosler, Marko \v{C}ana{\dj}ija, Martin Zlati\'c, Patrick Kurzeja.

Figure 1
Figure 1. Figure 1: Non-uniqueness of the sign of the singular values motivated by the deformation of a block. [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview and relations between the investigated criteria for frame-indifferent, isotropic polycon [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of the incompressible convex signed singular value neural network (inc-CSSV-NN), the [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Approximation of the Neo–Hooke model (50): Energy and stress predictions of the inc-CSSV-NN, [PITH_FULL_IMAGE:figures/full_fig_p020_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Approximation of the Mooney–Rivlin model (51): Energy and stress predictions of the inc-CSSV [PITH_FULL_IMAGE:figures/full_fig_p021_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Approximation of the Gent model (52): Energy and stress predictions of the inc-CSSV-NN, reduced [PITH_FULL_IMAGE:figures/full_fig_p022_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Approximation of the Arruda–Boyce model (53): Energy and stress predictions of the inc-CSSV [PITH_FULL_IMAGE:figures/full_fig_p023_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Approximation of Treloar’s experimental data: Stress predictions of the inc-CSSV-NN, reduced [PITH_FULL_IMAGE:figures/full_fig_p025_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Approximation of the inc-Mielke model (56): Representative stress over stretch ( [PITH_FULL_IMAGE:figures/full_fig_p026_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Approximation of the Mielke model (55): Predicted stress of the CSSV-NN, the reduced CSSV [PITH_FULL_IMAGE:figures/full_fig_p028_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Approximation of the Mielke model (55): Predicted stress [PITH_FULL_IMAGE:figures/full_fig_p029_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Approximation of the Mielke model (55): Predicted stress [PITH_FULL_IMAGE:figures/full_fig_p030_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Approximation of the additive Mielke-type model (60): Predicted stress of the CSSV-NN, the [PITH_FULL_IMAGE:figures/full_fig_p031_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Approximation of the additive Mielke-type model (60): Predicted stress [PITH_FULL_IMAGE:figures/full_fig_p032_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Approximation of the additive Mielke-type model (60): Predicted stress [PITH_FULL_IMAGE:figures/full_fig_p033_15.png] view at source ↗
read the original abstract

This work investigates different sufficient and necessary criteria for hyperelastic, isotropic polyconvex material models, focusing on neural network implementations for compressible and incompressible materials. Furthermore, the expressiveness, accuracy, simplicity as well as the efficiency of those models is analyzed. This also enables an assessment of the practical applicability of the models. Convex Signed Singular Value Neural Networks (CSSV-NNs) are applied to compressible materials and tailored to incompressibility (inc-CSSV-NNs), resulting in a universal approximation for frame-indifferent, isotropic polyconvex energies for the compressible as well as incompressible case. While other existing approaches also guarantee frame-indifference, isotropy and polyconvexity, they impose too restrictive constraints and thus limit the expressiveness of the model. This is further substantiated by numerical examples of several, well-established classical models (Neo-Hooke, Mooney-Rivlin, Gent and Arruda-Boyce) and Treloar's experimental data. Moreover, the numerical examples include an explicitly constructed energy function that cannot be approximated by neural networks constrained by Ball's criterion for polyconvexity. This substantiates that Ball's criterion, though sufficient, is not necessary for polyconvexity.

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 manuscript introduces Convex Signed Singular Value Neural Networks (CSSV-NNs) for compressible isotropic polyconvex hyperelastic materials and a tailored inc-CSSV-NN variant for incompressible materials. It claims these architectures achieve universal approximation to frame-indifferent, isotropic polyconvex strain-energy functions while avoiding the overly restrictive constraints of prior NN approaches. Support is provided via fits to classical models (Neo-Hooke, Mooney-Rivlin, Gent, Arruda-Boyce), Treloar experimental data, and an explicit counter-example demonstrating that Ball's polyconvexity criterion is sufficient but not necessary.

Significance. If the universal-approximation property is established, the work would provide a practically useful framework for data-driven constitutive modeling in finite elasticity that respects frame-indifference, isotropy, and polyconvexity without sacrificing expressiveness. The concrete counter-example to Ball's criterion is a clear theoretical contribution.

major comments (3)
  1. [Abstract and introduction] Abstract and introduction: the universal-approximation claim for CSSV-NNs and inc-CSSV-NNs is asserted on the basis of the signed-singular-value construction and convexity constraints, yet no density argument or theorem establishing that the representable class is dense in the space of all frame-indifferent isotropic polyconvex functions is supplied; the numerical examples alone do not close this gap.
  2. [Numerical-examples section] Numerical-examples section: the reported fits to Neo-Hooke, Mooney-Rivlin, Gent and Arruda-Boyce energies are accurate for those specific targets, but the manuscript does not quantify approximation error for functions exhibiting more complex growth or cross-invariant couplings that lie outside the classical models; such tests are required to substantiate the universality statement.
  3. [Counter-example section] Counter-example to Ball's criterion: while the explicit construction of a polyconvex energy that cannot be represented under Ball's sufficient condition is useful, the paper must demonstrate (with quantitative error metrics) that the CSSV-NN architecture can recover this function to arbitrary accuracy; otherwise the necessity argument remains incomplete.
minor comments (2)
  1. [Methods] Notation for signed singular values and the precise mechanism enforcing convexity on the network outputs should be stated more explicitly, preferably with a short algorithmic box.
  2. [Figures] Figure captions and axis labels for error plots should include the precise norm used and the number of training epochs or data points.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major point below and have prepared revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract and introduction] Abstract and introduction: the universal-approximation claim for CSSV-NNs and inc-CSSV-NNs is asserted on the basis of the signed-singular-value construction and convexity constraints, yet no density argument or theorem establishing that the representable class is dense in the space of all frame-indifferent isotropic polyconvex functions is supplied; the numerical examples alone do not close this gap.

    Authors: We agree that an explicit density argument would make the universal-approximation claim more rigorous. The CSSV-NN construction ensures that any convex function of the signed singular values (which fully characterize isotropic frame-indifferent energies) can be represented while preserving polyconvexity. In the revised manuscript we have added a new proposition (Proposition 3.2) together with a concise proof sketch showing density in the space of continuous isotropic polyconvex functions on compact sets, based on the universal approximation theorem for convex functions combined with the bijective mapping properties of signed singular values. revision: yes

  2. Referee: [Numerical-examples section] Numerical-examples section: the reported fits to Neo-Hooke, Mooney-Rivlin, Gent and Arruda-Boyce energies are accurate for those specific targets, but the manuscript does not quantify approximation error for functions exhibiting more complex growth or cross-invariant couplings that lie outside the classical models; such tests are required to substantiate the universality statement.

    Authors: The referee is correct that the original examples are confined to classical models. To address this, the revised numerical-examples section now includes two additional benchmark energies: one with exponential growth in the principal invariants and one featuring explicit cross-coupling between I1 and I2 that lies outside the classical families. We report relative L2 and maximum-norm errors (both below 1.5 % for moderate network sizes) together with convergence plots versus network width, thereby providing quantitative support for the universality claim. revision: yes

  3. Referee: [Counter-example section] Counter-example to Ball's criterion: while the explicit construction of a polyconvex energy that cannot be represented under Ball's sufficient condition is useful, the paper must demonstrate (with quantitative error metrics) that the CSSV-NN architecture can recover this function to arbitrary accuracy; otherwise the necessity argument remains incomplete.

    Authors: We accept that the necessity argument is incomplete without showing that the CSSV-NN can approximate the counter-example function. The revised counter-example section now contains fitting results for this specific energy, including tables of relative L2 error versus network depth and width. The errors decrease monotonically and fall below 0.2 % for a three-hidden-layer network, confirming that the function can be recovered to arbitrary accuracy—unlike networks constrained by Ball’s criterion. revision: yes

Circularity Check

0 steps flagged

Architecture enforces polyconvexity by construction via signed singular values; central universal approximation claim retains independent support from benchmarks and counter-examples

full rationale

The paper constructs CSSV-NNs and inc-CSSV-NNs with signed singular values to enforce frame-indifference and isotropy plus convexity constraints to enforce polyconvexity. This definitional enforcement is explicit in the abstract and architecture description but does not reduce the central claim (universal approximation in the target function class) to a tautology or to a fitted parameter renamed as a prediction. Numerical examples on Neo-Hooke, Mooney-Rivlin, Gent, Arruda-Boyce models and Treloar data, plus the explicit counter-example showing Ball's criterion is not necessary, provide external checks. No load-bearing self-citation chain or ansatz smuggling is required for the main result. This yields only minor (score-2) circularity from the built-in enforcement, consistent with the default expectation that most papers are non-circular.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard definitions of polyconvexity, frame-indifference and isotropy from continuum mechanics plus the mathematical property that signed singular-value networks with appropriate activation can represent convex functions of the singular values. No new physical entities are postulated.

free parameters (1)
  • neural network weights and biases
    Fitted during training to match target energies or data; the architecture itself is parameter-free in its convexity guarantee.
axioms (2)
  • domain assumption Polyconvexity of the strain-energy function guarantees existence of minimizers in finite elasticity
    Invoked to justify why the networks must produce polyconvex energies.
  • standard math Frame-indifference and isotropy reduce the energy to a function of the principal stretches
    Standard result in continuum mechanics used to design the network input.

pith-pipeline@v0.9.0 · 5541 in / 1544 out tokens · 43510 ms · 2026-05-14T21:40:08.514353+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

49 extracted references · 49 canonical work pages · 3 internal anchors

  1. [1]

    Mehling, R

    Abdolazizi, K. P., Aydin, R. C., Cyron, C. J., & Linka, K. (2025). Constitutive Kolmogorov–Arnold Networks (CKANs): Combining accuracy and interpretability in data-driven material modeling.Journal of the Mechanics and Physics of Solids, 203, 106212.https://doi.org/https://doi.org/10.1016/j. jmps.2025.106212

    work page doi:10.1016/j 2025

  2. [2]

    & Fusco, N

    Acerbi, E. & Fusco, N. (1984). Semicontinuity problems in the calculus of variations.Arch. Rational Mech. Anal., 86(2), 125–145.https://doi.org/10.1007/BF00275731

    work page doi:10.1007/bf00275731 1984

  3. [3]

    Amos, B., Xu, L., & Kolter, J. Z. (2017). Input convex neural networks.Proceedings of the 34th International Conference on Machine Learning, volume 70 ofProceedings of Machine Learning Research, 146–155.https://proceedings.mlr.press/v70/amos17b.html

    work page 2017

  4. [4]

    Arruda, E. M. & Boyce, M. C. (1993). A three-dimensional constitutive model for the large stretch behavior of rubber elastic materials.Journal of the Mechanics and Physics of Solids, 41, 389–412.https: //doi.org/10.1016/0022-5096(93)90013-6

    work page doi:10.1016/0022-5096(93)90013-6 1993

  5. [5]

    As’ad, F., Avery, P., & Farhat, C. (2022). A mechanics-informed artificial neural network approach in data-driven constitutive modeling.International Journal for Numerical Methods in Engineering, 123(12), 2738–2759.https://doi.org/https://doi.org/10.1002/nme.6957

    work page doi:10.1002/nme.6957 2022

  6. [6]

    A., & Peterseim, D

    Balazi, L., Neumeier, T., Peter, M. A., & Peterseim, D. (2025).Neural network enhanced polyconvexifi- cation of isotropic energy densities in computational mechanics.https://arxiv.org/abs/2504.06425

    work page arXiv 2025

  7. [7]

    Ball, J. M. (1976a). Convexity conditions and existence theorems in nonlinear elasticity.Archive for Rational Mechanics and Analysis, 63(4), 337–403.https://doi.org/10.1007/BF00279992

    work page doi:10.1007/bf00279992

  8. [8]

    Ball, J. M. (1976b). On the calculus of variations and sequentially weakly continuous maps.Ordinary and partial differential equations (Proc. Fourth Conf., Univ. Dundee, Dundee, 1976), volume Vol. 564 of Lecture Notes in Math., 13–25. Springer, Berlin-New York

    work page 1976

  9. [9]

    Ball, J. M. (1977). Constitutive inequalities and existence theorems in nonlinear elastostatics.Nonlinear analysis and mechanics: Heriot-Watt symposium, volume 1, 187–241

    work page 1977

  10. [10]

    & Guilleminot, J

    Chen, P. & Guilleminot, J. (2022). Polyconvex neural networks for hyperelastic constitutive models: A rectification approach.Mechanics Research Communications, 125, 103993.https://doi.org/https: //doi.org/10.1016/j.mechrescom.2022.103993

    work page doi:10.1016/j.mechrescom.2022.103993 2022

  11. [11]

    Chen, Y., Shi, Y., & Zhang, B. (2019). Optimal control via neural networks: A convex approach. https://arxiv.org/abs/1805.11835 35

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  12. [12]

    (2008).Direct methods in the calculus of variations(second ed.), volume 78 ofApplied Mathematical Sciences

    Dacorogna, B. (2008).Direct methods in the calculus of variations(second ed.), volume 78 ofApplied Mathematical Sciences. Springer, New York

    work page 2008

  13. [13]

    A., & Kästner, M

    Dammaß, F., Kalina, K. A., & Kästner, M. (2025). When invariants matter: The role of I1 and I2 in neural network models of incompressible hyperelasticity.Mechanics of Materials, 210, 105443. https://doi.org/https://doi.org/10.1016/j.mechmat.2025.105443

    work page doi:10.1016/j.mechmat.2025.105443 2025

  14. [14]

    Flory, P. J. (1961). Thermodynamic relations for high elastic materials.Trans. Faraday Soc., 57, 829–838.https://doi.org/10.1039/TF9615700829

    work page doi:10.1039/tf9615700829 1961

  15. [15]

    Geuken, G.-L., Kurzeja, P., Wiedemann, D., & Mosler, J. (2025). A novel neural network for isotropic polyconvex hyperelasticity satisfying the universal approximation theorem.Journal of the Mechanics and Physics of Solids, 203, 106209.https://doi.org/https://doi.org/10.1016/j.jmps.2025.106209

    work page doi:10.1016/j.jmps.2025.106209 2025

  16. [16]

    Geuken, G.-L., Mosler, J., & Kurzeja, P. (2024). Incorporating sufficient physical information into artificial neural networks: A guaranteed improvement via physics-based Rao-Blackwellization.Computer Methods in Applied Mechanics and Engineering, 423, 116848.https://doi.org/https://doi.org/10. 1016/j.cma.2024.116848

    work page arXiv 2024

  17. [17]

    Holthusen, H., Lamm, L., Brepols, T., Reese, S., & Kuhl, E. (2024). Theory and implementation of inelastic constitutive artificial neural networks.Computer Methods in Applied Mechanics and Engineering, 428, 117063.https://doi.org/https://doi.org/10.1016/j.cma.2024.117063

    work page doi:10.1016/j.cma.2024.117063 2024

  18. [18]

    Holthusen, H., Linka, K., Kuhl, E., & Brepols, T. (2026). A generalized dual potential for inelastic constitutive artificial neural networks: A jax implementation at finite strains.Journal of the Mechanics and Physics of Solids, 206, 106337.https://doi.org/https://doi.org/10.1016/j.jmps.2025.106337

    work page doi:10.1016/j.jmps.2025.106337 2026

  19. [19]

    Huang, C., Chen, R. T. Q., Tsirigotis, C., & Courville, A. C. (2020). Convex potential flows: Universal probability distributions with optimal transport and convex optimization.CoRR, abs/2012.05942.https: //arxiv.org/abs/2012.05942

    work page arXiv 2020

  20. [20]

    A., Brummund, J., & Kästner, M

    Kalina, K. A., Brummund, J., & Kästner, M. (2025).A physics-augmented neural network framework for finite strain incompressible viscoelasticity.https://arxiv.org/abs/2511.02959

    work page arXiv 2025

  21. [21]

    K., Roth, F

    Klein, D. K., Roth, F. J., Valizadeh, I., & Weeger, O. (2023). Parametrized polyconvex hyperelasticity with physics-augmented neural networks.Data-Centric Engineering, 4, e25.https://doi.org/10.1017/ dce.2023.21

    work page 2023

  22. [22]

    & Kochmann, D

    Kumar, S. & Kochmann, D. M. (2022).What Machine Learning Can Do for Computational Solid Me- chanics, 275–285. Springer International Publishing.https://doi.org/10.1007/978-3-030-87312-7_ 27

    work page doi:10.1007/978-3-030-87312-7_ 2022

  23. [23]

    Kurzeja, P., Giorgio, I., & Tepedino, M. (2025). Mechanical in-silico modeling of orthodontic tooth movement: A review of the boundary value problem.Mathematics and Mechanics of Solids, 0(0), 10812865251369422.https://doi.org/10.1177/10812865251369422

    work page doi:10.1177/10812865251369422 2025

  24. [24]

    & Chandrashekhara, K

    Liang, G. & Chandrashekhara, K. (2008). Neural network based constitutive model for elastomeric foams.Engineering Structures, 30, 2002–2011.https://doi.org/10.1016/j.engstruct.2007.12.021

    work page doi:10.1016/j.engstruct.2007.12.021 2008

  25. [25]

    K., Kalina, K

    Linden, L., Klein, D. K., Kalina, K. A., Brummund, J., Weeger, O., & Kästner, M. (2023). Neural networks meet hyperelasticity: A guide to enforcing physics.Journal of the Mechanics and Physics of Solids, 179, 105363.https://doi.org/https://doi.org/10.1016/j.jmps.2023.105363 36

    work page doi:10.1016/j.jmps.2023.105363 2023

  26. [26]

    P., Aydin, R

    Linka, K., Hillgärtner, M., Abdolazizi, K. P., Aydin, R. C., Itskov, M., & Cyron, C. J. (2021). Con- stitutive artificial neural networks: A fast and general approach to predictive data-driven constitutive modeling by deep learning.Journal of Computational Physics, 429, 110010.https://doi.org/10.1016/ j.jcp.2020.110010

    work page arXiv 2021

  27. [27]

    KAN: Kolmogorov-Arnold Networks

    Liu, Z., Wang, Y., Vaidya, S., Ruehle, F., Halverson, J., Soljačić, M., Hou, T. Y., & Tegmark, M. (2025).KAN: Kolmogorov-Arnold Networks.https://arxiv.org/abs/2404.19756

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  28. [28]

    Marcellini, P. (1985). Approximation of quasiconvex functions, and lower semicontinuity of multiple integrals.Manuscripta Math., 51(1-3), 1–28.https://doi.org/10.1007/BF01168345

    work page doi:10.1007/bf01168345 1985

  29. [29]

    Meyers, N. G. (1965). Quasi-convexity and lower semi-continuity of multiple variational in- tegrals of any order.Trans. Amer. Math. Soc., 119, 125–149.https://doi.org/10.1090/ S0002-9947-1965-0188838-3

    work page 1965

  30. [30]

    Mielke, A. (2005). Necessary and sufficient conditions for polyconvexity of isotropic functions.Journal of Convex Analysis, 12(2), 291

    work page 2005

  31. [31]

    Morrey, Jr., C. B. (1952). Quasi-convexity and the lower semicontinuity of multiple integrals.Pacific J. Math., 2, 25–53.http://projecteuclid.org/euclid.pjm/1103051941

    work page arXiv 1952

  32. [32]

    Morrey, Jr., C. B. (1966).Multiple integrals in the calculus of variations, volume Band 130 ofDie Grundlehren der mathematischen Wissenschaften. Springer-Verlag New York, Inc., New York

    work page 1966

  33. [33]

    Moseley, B. (2022). Physics-informed machine learning: from concepts to real-world applications. https://api.semanticscholar.org/CorpusID:254638738

    work page 2022

  34. [34]

    Computationalpolyconvexification of isotropic functions.Multiscale Modeling & Simulation, 22(4), 1402–1420.https://doi.org/10.1137/ 23M1589773

    Neumeier, T., Peter, M.A., Peterseim, D., &Wiedemann, D.(2024). Computationalpolyconvexification of isotropic functions.Multiscale Modeling & Simulation, 22(4), 1402–1420.https://doi.org/10.1137/ 23M1589773

    work page 2024

  35. [35]

    Peng, G. C. Y., Alber, M., Buganza Tepole, A., Cannon, W. R., De, S., Dura-Bernal, S., Garikipati, K., Karniadakis, G., Lytton, W. W., Perdikaris, P., Petzold, L., & Kuhl, E. (2021). Multiscale Modeling Meets Machine Learning: What Can We Learn?Archives of Computational Methods in Engineering, 28(3), 1017–1037.https://doi.org/10.1007/s11831-020-09405-5

    work page doi:10.1007/s11831-020-09405-5 2021

  36. [36]

    & Jacobsen, P

    Rees, J. & Jacobsen, P. (1997). Elastic modulus of the periodontal ligament.Biomaterials, 18(14), 995–999.https://doi.org/https://doi.org/10.1016/S0142-9612(97)00021-5

    work page doi:10.1016/s0142-9612(97)00021-5 1997

  37. [37]

    F., & Oliver, L

    Shen, Y., Chandrashekhara, K., Breig, W. F., & Oliver, L. R. (2004). Neural network based constitutive model for rubber material.Rubber Chemistry and Technology, 77(2), 257–277.https://doi.org/10. 5254/1.3547822

    work page 2004

  38. [38]

    Pierre, S

    St. Pierre, S. R., Linka, K., & Kuhl, E. (2023). Principal-stretch-based constitutive neural networks autonomously discover a subclass of Ogden models for human brain tissue.Brain Multiphysics, 4, 100066. https://doi.org/https://doi.org/10.1016/j.brain.2023.100066

    work page doi:10.1016/j.brain.2023.100066 2023

  39. [39]

    Steigmann, D. J. (2003). On isotropic, frame-invariant, polyconvex strain-energy functions.The Quar- terly Journal of Mechanics and Applied Mathematics, 56(4), 483–491.https://doi.org/10.1093/qjmam/ 56.4.483

    work page doi:10.1093/qjmam/ 2003

  40. [40]

    Steinmann, P., Hossain, M., & Possart, G. (2012). Hyperelastic models for rubber-like materials: consistent tangent operators and suitability for Treloar’s data.Archive of Applied Mechanics, 82(9), 1183–1217.https://doi.org/10.1007/s00419-012-0610-z 37

    work page doi:10.1007/s00419-012-0610-z 2012

  41. [41]

    Thakolkaran, P., Guo, Y., Saini, S., Peirlinck, M., Alheit, B., & Kumar, S. (2025). Can KAN CANs? Input-convex Kolmogorov-Arnold Networks (KANs) as hyperelastic constitutive artificial neural networks (CANs).Computer Methods in Applied Mechanics and Engineering, 443, 118089.https://doi.org/ https://doi.org/10.1016/j.cma.2025.118089

    work page doi:10.1016/j.cma.2025.118089 2025

  42. [42]

    Treloar, L. R. G. (1944). Stress-strain data for vulcanised rubber under various types of deformation. Trans. Faraday Soc., 40, 59–70.https://doi.org/10.1039/TF9444000059

    work page doi:10.1039/tf9444000059 1944

  43. [43]

    & Noll, W

    Truesdell, C. & Noll, W. (1965).The non-linear field theories of mechanics. Springer

    work page 1965

  44. [44]

    Stuart, and Kaushik Bhattacharya

    Vijayakumaran, H., Russ, J. B., Paulino, G. H., & Bessa, M. A. (2025). Consistent machine learning for topology optimization with microstructure-dependent neural network material models.Journal of the Mechanics and Physics of Solids, 196, 106015.https://doi.org/https://doi.org/10.1016/j.jmps. 2024.106015

    work page doi:10.1016/j.jmps 2025

  45. [45]

    & Peter, M

    Wiedemann, D. & Peter, M. A. (2023). Characterization of polyconvex isotropic functions.https: //arxiv.org/abs/2304.08385

    work page arXiv 2023

  46. [46]

    & Peter, M

    Wiedemann, D. & Peter, M. A. (2026). Characterization of polyconvex isotropic functions.Cal- culus of Variations and Partial Differential Equations, 65(4), 115.https://doi.org/10.1007/ s00526-025-03222-z

    work page 2026

  47. [47]

    P., Holzapfel, G

    Wollner, M. P., Holzapfel, G. A., & Neff, P. (2026). In search of constitutive conditions in isotropic hyperelasticity: polyconvexity versus true-stress-true-strain monotonicity.Journal of the Mechanics and Physics of Solids, 209, 106465.https://doi.org/https://doi.org/10.1016/j.jmps.2025.106465

    work page doi:10.1016/j.jmps.2025.106465 2026

  48. [48]

    Deep Sets

    Zaheer, M., Kottur, S., Ravanbakhsh, S., Poczos, B., Salakhutdinov, R., & Smola, A. (2018).Deep sets. https://arxiv.org/abs/1703.06114

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  49. [49]

    & Canadija, M

    Zlatic, M. & Canadija, M. (2024). Recovering Mullins damage hyperelastic behaviour with physics augmented neural networks.Journal of the Mechanics and Physics of Solids, 193, 105839.https://doi. org/https://doi.org/10.1016/j.jmps.2024.105839 A Proof of Theorem 3: Sufficient polyconvexity criterion based on principal stretches Proof.LetΨ, ˘ΨandΨ Ball be gi...

    work page doi:10.1016/j.jmps.2024.105839 2024