pith. sign in

arxiv: 2605.20816 · v1 · pith:IM55VDAZnew · submitted 2026-05-20 · 🧮 math.OC · cond-mat.mtrl-sci

Polynomial diagrams for microstructure modelling

David P. Bourne , Maciej Buze , Thomas Gallou\"et , Quentin M\'erigot This is my paper

Pith reviewed 2026-05-21 03:57 UTC · model grok-4.3

classification 🧮 math.OC cond-mat.mtrl-sci
keywords polynomial diagramspower diagramsmicrostructure modellingimage segmentationLegendre polynomialsconvex optimizationmaterials scienceelectron backscatter diffraction
0
0 comments X

The pith

Polynomial diagrams generalize power diagrams so cell boundaries become algebraic curves of any chosen degree.

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

This paper develops a framework for polynomial diagrams that extend power diagrams and anisotropic power diagrams by permitting boundaries to be algebraic curves of any chosen degree. These diagrams arise by rephrasing standard diagrams as higher-degree cases of linear parametrised minimisation diagrams. The authors supply a GPU-accelerated fitting procedure that uses Legendre polynomials to maximise a regularised concave objective adapted from logistic regression. The method is applied to electron backscatter diffraction images of steel microstructures. A sympathetic reader would care because the extra flexibility lets models capture more complex grain shapes that appear in real materials.

Core claim

We formulate a framework of polynomial diagrams, which are a generalisation of power diagrams (PDs) and anisotropic power diagrams (APDs) allowing for boundaries between cells to be algebraic curves of a prescribed degree. We show that they arise naturally from rephrasing PDs (APDs) as first-degree (second-degree) instances of linear parametrised minimisation diagrams. We also develop an efficient GPU-accelerated framework for fitting polynomial diagrams to image data using Legendre polynomials and by maximising a regularised concave objective function adapted from classical logistic regression literature.

What carries the argument

Polynomial diagrams defined as instances of linear parametrised minimisation diagrams, which produce algebraic boundaries of any prescribed degree between cells.

Load-bearing premise

Microstructures in electron backscatter diffraction images of steel can be meaningfully represented and fitted by polynomial diagrams of a chosen degree arising from linear parametrised minimisation diagrams.

What would settle it

Fitting polynomial diagrams of degree three or higher to the same steel EBSD images produces no measurable improvement in boundary accuracy or objective value over degree-two anisotropic power diagrams.

Figures

Figures reproduced from arXiv: 2605.20816 by David P. Bourne, Maciej Buze, Quentin M\'erigot, Thomas Gallou\"et.

Figure 1
Figure 1. Figure 1: The reconstruction of (anisotropic) power diagrams as described in Section 5.4. [PITH_FULL_IMAGE:figures/full_fig_p025_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The fitting of polynomial diagrams to the small EBSD dataset as described in [PITH_FULL_IMAGE:figures/full_fig_p026_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The fitting of polynomial diagrams to the big EBSD dataset as described in Sec [PITH_FULL_IMAGE:figures/full_fig_p027_3.png] view at source ↗
read the original abstract

We formulate a framework of polynomial diagrams, which are a generalisation of power diagrams (PDs) and anisotropic power diagrams (APDs) allowing for boundaries between cells to be algebraic curves of a prescribed degree. We show that they arise naturally from rephrasing PDs (APDs) as first-degree (second-degree) instances of linear parametrised minimisation diagrams. We also develop an efficient GPU-accelerated framework for fitting polynomial diagrams to image data using Legendre polynomials and by maximising a regularised concave objective function adapted from classical logistic regression literature. A largely self-contained analysis of the optimisation algorithm is also provided, including identification of scale and gauge invariances and the limiting objective function as the regularisation parameter vanishes. We apply the algorithm to fit polynomial diagrams to electron backscatter diffraction images of steel.

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

2 major / 2 minor

Summary. The paper formulates polynomial diagrams as a generalization of power diagrams (PDs) and anisotropic power diagrams (APDs) in which cell boundaries are algebraic curves of a prescribed degree. These arise from rephrasing PDs and APDs as degree-1 and degree-2 cases of linear parametrised minimisation diagrams, extended via Legendre parametrisation. An efficient GPU-accelerated fitting procedure is developed that maximises a regularised concave objective adapted from logistic regression. The optimisation algorithm is analysed for scale and gauge invariances and the limiting objective as the regularisation parameter tends to zero. The method is applied to fitting electron backscatter diffraction (EBSD) images of steel microstructures.

Significance. If the central claims hold, the framework supplies a flexible, parametrised family of diagrams with algebraic boundaries of controllable degree together with a practical GPU fitting pipeline and a largely self-contained convergence analysis. These elements could be useful for microstructure modelling where grain boundaries deviate from the linear or quadratic cases already covered by PDs and APDs. The explicit treatment of invariances and the limiting objective is a clear strength that aids reproducibility and theoretical understanding.

major comments (2)
  1. [Application to EBSD images] Application section (EBSD steel images): the manuscript shows visual fits but supplies no quantitative metrics (pixel-wise error, Hausdorff distance on boundaries, or cross-validation scores) comparing polynomial diagrams of degree >2 against the APD (degree-2) baseline. Without such numbers or an ablation on degree and regularisation strength, it remains unclear whether the additional degrees reduce representation error on real data or merely increase parameter count.
  2. [Framework formulation] Framework formulation: the claim that boundaries are exactly algebraic curves of the prescribed degree relies on the Legendre parametrisation of the linear minimisation diagram; the manuscript should explicitly verify that the chosen basis and truncation enforce this degree exactly rather than approximately, especially when the regularised objective is maximised.
minor comments (2)
  1. [Notation and definitions] Clarify the precise definition and notation for 'linear parametrised minimisation diagrams' at first use, including how the parameters enter the minimisation.
  2. [Results figures] Figure captions in the results should state the polynomial degree, regularisation value, and data resolution for each displayed fit.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments on the manuscript. We address each major comment below and have revised the paper accordingly to strengthen the presentation.

read point-by-point responses

  1. Referee: [Application to EBSD images] Application section (EBSD steel images): the manuscript shows visual fits but supplies no quantitative metrics (pixel-wise error, Hausdorff distance on boundaries, or cross-validation scores) comparing polynomial diagrams of degree >2 against the APD (degree-2) baseline. Without such numbers or an ablation on degree and regularisation strength, it remains unclear whether the additional degrees reduce representation error on real data or merely increase parameter count.

    Authors: We agree that quantitative metrics would provide stronger support for the utility of higher-degree diagrams. The original submission emphasised visual results to illustrate the method on EBSD data. In the revised manuscript we have added a quantitative comparison subsection that reports pixel-wise misclassification error and boundary Hausdorff distances for degrees 1, 2 and 3, together with an ablation table over degree and regularisation strength. These results indicate a modest but consistent improvement for degree 3 over the APD baseline on the steel images, with regularisation controlling parameter growth. revision: yes

  2. Referee: [Framework formulation] Framework formulation: the claim that boundaries are exactly algebraic curves of the prescribed degree relies on the Legendre parametrisation of the linear minimisation diagram; the manuscript should explicitly verify that the chosen basis and truncation enforce this degree exactly rather than approximately, especially when the regularised objective is maximised.

    Authors: The Legendre basis of order at most d spans all polynomials of degree ≤ d, so the resulting minimisation diagram yields cell boundaries that are exactly the zero loci of degree-d polynomials (i.e., algebraic curves of exact degree d). The regularisation term modifies only the objective landscape and does not change the functional form or degree of the boundaries. We have inserted a short clarifying remark and verification argument in Section 2.2 of the revised manuscript to make this explicit. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper introduces polynomial diagrams as a mathematical generalization of power diagrams (PDs) and anisotropic power diagrams (APDs) by rephrasing the latter as degree-1 and degree-2 cases of linear parametrised minimisation diagrams, then extends the construction to higher-degree algebraic boundaries via Legendre parametrisation. It develops a GPU-accelerated fitting procedure that maximises a regularised concave objective adapted from external logistic regression literature and supplies a self-contained analysis of scale/gauge invariances and the vanishing-regularisation limit. The method is applied to independent EBSD steel image data. No derivation step reduces a claimed prediction or result to a fitted parameter or prior self-citation by construction; all load-bearing components either rest on external literature or on direct fitting to outside data.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 1 invented entities

The central claim rests on the mathematical rephrasing of power diagrams into a general minimization framework and the assumption that image data from steel microstructures can be fitted effectively with the proposed optimization.

free parameters (2)
  • polynomial degree
    Prescribed degree for the algebraic curves defining cell boundaries in the diagrams.
  • regularization parameter
    Parameter in the concave objective function whose vanishing limit is analyzed in the optimization study.
axioms (1)
  • domain assumption Polynomial diagrams arise naturally from rephrasing PDs and APDs as first-degree and second-degree instances of linear parametrised minimisation diagrams.
    Core formulation step stated in the abstract for the framework.
invented entities (1)
  • polynomial diagram no independent evidence
    purpose: Generalization allowing algebraic curve boundaries of prescribed degree for microstructure cell modeling.
    New framework introduced to extend power diagrams.

pith-pipeline@v0.9.0 · 5669 in / 1436 out tokens · 50898 ms · 2026-05-21T03:57:25.648728+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

55 extracted references · 55 canonical work pages

  1. [1]

    Sparsistency for Inverse Optimal Transport , booktitle =

    Francisco Andrade and Gabriel Peyr. Sparsistency for Inverse Optimal Transport , booktitle =

    work page

  2. [2]

    2025 , note =

    Clarice Poon , title =. 2025 , note =

    work page 2025

  3. [3]

    International Conference on Machine Learning , pages=

    Discrete probabilistic inverse optimal transport , author=. International Conference on Machine Learning , pages=. 2022 , organization=

    work page 2022

  4. [4]

    Journal of Machine Learning Research , volume =

    Ruilin Li and Xiaojing Ye and Haomin Zhou and Hongyuan Zha , title =. Journal of Machine Learning Research , volume =. 2019 , pages =

    work page 2019

  5. [5]

    Asymptotic and computational analysis , pages=

    How (un) stable are Vandermonde systems? , author=. Asymptotic and computational analysis , pages=. 2020 , publisher=

    work page 2020

  6. [6]

    2019 , publisher=

    Approximation theory and approximation practice, extended edition , author=. 2019 , publisher=

    work page 2019

  7. [7]

    and Brereton, T

    Spettl, A. and Brereton, T. and Duan, Q. and Werz, T. and. Fitting. Philosophical Magazine , volume =. 2016 , doi =

    work page 2016

  8. [8]

    and Kubler, R

    Depriester, D. and Kubler, R. , TITLE =. Computer-Aided Design , VOLUME =. 2019 , PAGES =

    work page 2019

  9. [9]

    and Schneider, M

    Kuhn, J. and Schneider, M. and Sonnweber-Ribic, P. and B. Fast methods for computing centroidal. Computer Methods in Applied Mechanics and Engineering , volume=. 2020 , doi=

    work page 2020

  10. [10]

    and Steinhauser, M

    K\"uhn, M. and Steinhauser, M. O. , title =. Applied Physics Letters , volume =. 2008 , doi =

    work page 2008

  11. [11]

    Philosophical Magazine , volume=

    Laguerre tessellations and polycrystalline microstructures: a fast algorithm for generating grains of given volumes , author=. Philosophical Magazine , volume=. 2020 , doi=

    work page 2020

  12. [12]

    Mechanics Research Communications , volume=

    Geometric modelling of polycrystalline materials: Laguerre tessellations and periodic semi-discrete optimal transport , author=. Mechanics Research Communications , volume=. 2023 , doi =

    work page 2023

  13. [13]

    Bourne and Thomas O

    David P. Bourne and Thomas O. Gallou\"et and Quentin M\'erigot and Andrea Natale , title =. hal-05008630 , year =

    work page

  14. [14]

    Bourne and Mason Pearce and Steven M

    David P. Bourne and Mason Pearce and Steven M. Roper , title =. ESAIM: Mathematical Modelling and Numerical Analysis , volume =. 2025 , doi =

    work page 2025

  15. [15]

    Microscopy and Microanalysis , author=

    Reconstruction of Grains in Polycrystalline Materials From Incomplete Data Using. Microscopy and Microanalysis , author=. 2019 , pages=

    work page 2019

  16. [16]

    Alpers and O

    A. Alpers and O. Furat and C. Jung and M. Neumann and C. Redenbach and A. Saken and V. Schmidt , title =. 2025 , journal =

    work page 2025

  17. [17]

    Computer-Aided Design , doi =

    Jocelyn Meyron , title =. Computer-Aided Design , doi =. 2019 , volume =

    work page 2019

  18. [18]

    In Search of Empty Spheres:

    Plateau-Holleville, Cyprien and Stamm, Benjamin and Nivoliers, Vincent and Maria, Maxime and M. In Search of Empty Spheres:. ACM Transactions on Graphics (TOG) , volume=. 2025 , publisher=

    work page 2025

  19. [19]

    Computer Methods in Applied Mechanics and Engineering , volume=

    Generating high-fidelity microstructures of polycrystalline materials with prescribed higher-order texture tensors , author=. Computer Methods in Applied Mechanics and Engineering , volume=. 2026 , publisher=

    work page 2026

  20. [20]

    and Raabe, Dierk , title =

    Roters, Franz and Eisenlohr, Philip and Hantcherli, Luc and Tjahjanto, Denny Dharmawan and Bieler, Thomas R. and Raabe, Dierk , title =. Acta Materialia , year =

    work page

  21. [21]

    Anderson, M. P. and Srolovitz, D. J. and Grest, G. S. and Sahni, P. S. , title =. Acta Metallurgica , year =

    work page

  22. [22]

    and Kocks, U

    Mecking, H. and Kocks, U. F. , title =. Acta Metallurgica , year =

    work page

  23. [23]

    Calphad , year =

    Moelans, Nele and Blanpain, Bart and Wollants, Patrick , title =. Calphad , year =

    work page

  24. [24]

    Hall, E. O. , title =. Proceedings of the Physical Society. Section B , year =

    work page

  25. [25]

    Optimal polyhedral description of

    Quey, Romain and Renversade, Lo. Optimal polyhedral description of. Computer Methods in Applied Mechanics and Engineering , year =

    work page

  26. [26]

    Advanced Engineering Materials , year =

    Lyckegaard, Allan and Lauridsen, Erik Mejdal and Ludwig, Wolfgang and Fonda, Richard Warren and Poulsen, Henning Friis , title =. Advanced Engineering Materials , year =

    work page

  27. [27]

    Computational Materials Science , year =

    Fan, Zhigang and Wu, Yugong and Zhao, Xuanhe and Lu, Yuzhu , title =. Computational Materials Science , year =

    work page

  28. [28]

    Philosophical Magazine , year =

    Alpers, Andreas and Fiedler, Maximilian and Gritzmann, Peter and Klemm, Fabian , title =. Philosophical Magazine , year =

    work page

  29. [29]

    Philosophical Magazine , year =

    work page

  30. [30]

    2018 , volume =

    Data-driven selection of tessellation models describing polycrystalline microstructures , journal =. 2018 , volume =

    work page 2018

  31. [31]

    Image Analysis and Stereology , year =

    Altendorf, Hellen and Latourte, Felix and Jeulin, Dominique and Faessel, Matthieu and Saintoyant, Lucie , title =. Image Analysis and Stereology , year =

    work page

  32. [32]

    Computational Geometry , year =

    Jung, Christian and Redenbach, Claudia , title =. Computational Geometry , year =

    work page

  33. [33]

    Philosophical Magazine , year =

    Alpers, Andreas and Brieden, Andreas and Gritzmann, Peter and Lyckegaard, Allan and Poulsen, Henning Friis , title =. Philosophical Magazine , year =

    work page

  34. [34]

    Ultramicroscopy , year =

    Bachmann, Florian and Hielscher, Ralf and Schaeben, Helmut , title =. Ultramicroscopy , year =

    work page

  35. [35]

    van Laarhoven, Peter J. M. and Aarts, Emile H. L. , title =

    work page

  36. [36]

    and Nocedal, Jorge , title =

    Liu, Dong C. and Nocedal, Jorge , title =. Mathematical Programming , year =

    work page

  37. [37]

    Kernel operations on the

    Charlier, Benjamin and Feydy, Jean and Glaun. Kernel operations on the. Journal of Machine Learning Research , year =

    work page

  38. [38]

    and Feydy, J

    Buze, M. and Feydy, J. and Roper, S. M. and Sedighiani, K. and Bourne, D. P. , title =. Computational Materials Science , year =

    work page

  39. [39]

    2026 , howpublished =

    , title =. 2026 , howpublished =

    work page 2026

  40. [40]

    Algorithmica , year =

    Papadopoulou, Evanthia and Zavershynskyi, Maksym , title =. Algorithmica , year =

    work page

  41. [41]

    IEEE Transactions on Computers , year =

    Lee, Der-Tsai , title =. IEEE Transactions on Computers , year =

    work page

  42. [42]

    ACM Computing Surveys , year =

    Aurenhammer, Franz , title =. ACM Computing Surveys , year =

    work page

  43. [43]

    Journal of Machine Learning Research , year =

    Rifkin, Ryan and Klautau, Aldebaro , title =. Journal of Machine Learning Research , year =

    work page

  44. [44]

    and Higham, Nicholas J

    Blanchard, Pierre and Higham, Desmond J. and Higham, Nicholas J. , title =. IMA Journal of Numerical Analysis , year =

    work page

  45. [45]

    Boyd, Stephen and Vandenberghe, Lieven , title =

    work page

  46. [46]

    Hastie, Trevor and Tibshirani, Robert and Friedman, Jerome , title =

    work page

  47. [47]

    , title =

    Murphy, Kevin P. , title =

    work page

  48. [48]

    , title =

    Albert, Adelin and Anderson, John A. , title =. Biometrika , year =

    work page

  49. [49]

    and Duffy, Diane E

    Santner, Thomas J. and Duffy, Diane E. , title =. Biometrika , year =

    work page

  50. [50]

    2021 , howpublished =

    Feinman, Reuben , title =. 2021 , howpublished =

    work page 2021

  51. [51]

    Advances in Neural Information Processing Systems , year =

    Paszke, Adam and Gross, Sam and Massa, Francisco and Lerer, Adam and Bradbury, James and Chanan, Gregory and Killeen, Trevor and Lin, Zeming and Gimelshein, Natalia and Antiga, Luca and others , title =. Advances in Neural Information Processing Systems , year =

    work page

  52. [52]

    , title =

    Teferra, Kirubel and Rowenhorst, David J. , title =. Philosophical Magazine Letters , year =

    work page

  53. [53]

    SIAM Journal on Imaging Sciences , year =

    Alpers, Andreas and Fiedler, Maximilian and Gritzmann, Peter and Klemm, Fabian , title =. SIAM Journal on Imaging Sciences , year =

    work page

  54. [54]

    Computational optimal transport , journal =

    Peyr. Computational optimal transport , journal =. 2019 , volume =

    work page 2019

  55. [55]

    and Schmidt, Volker , title =

    Petrich, Lukas and Furat, Orkun and Wang, Mingyan and Krill III, Carl E. and Schmidt, Volker , title =. Frontiers in Materials , year =

    work page