Pith. sign in

REVIEW 3 major objections 2 minor 1 cited by

Neural networks approximate minimal surfaces in hyperbolic space whose self-intersections match Fine's conjecture for all tested knots

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-06-29 20:06 UTC pith:TUTWR6QF

load-bearing objection PINN tests of Fine's conjecture need validation on known surfaces before trusting the intersection counts. the 3 major comments →

arxiv 2605.26234 v2 pith:TUTWR6QF submitted 2026-05-25 math.DG cs.LGmath.GT

Minimal surfaces, Knots, and Neural Networks

Tancredi Schettini Gherardini , Marco Usula This is my paper
classification math.DG cs.LGmath.GT
keywords minimal surfacesknotsFine's conjecturephysics-informed neural networkshyperbolic 4-spaceHOMFLY polynomialself-intersections
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper develops a physics-informed neural network framework to solve the minimal surface equation in hyperbolic 4-space for surfaces that bound given knots at infinity. The method is applied to families of knots to produce approximate surfaces of prescribed genus and to extract their signed self-intersection numbers. These extracted numbers agree exactly with the values predicted by Fine's conjecture from the coefficients of the HOMFLY polynomial. A sympathetic reader would care because the work supplies the first numerical evidence linking a classical knot invariant to a geometric count in four-dimensional hyperbolic geometry.

Core claim

By training physics-informed neural networks to satisfy the minimal surface equation in H^4, we obtain near-minimal surfaces meeting the sphere at infinity along each tested knot K; for every such knot the signed count of self-intersections of these surfaces coincides with the count required by Fine's conjecture.

What carries the argument

A physics-informed neural network that minimises a loss encoding the minimal-surface PDE in hyperbolic 4-space, together with a post-processing algorithm that locates and signs the self-intersections of the resulting numerical surface.

Load-bearing premise

The surfaces produced by the neural networks are close enough to genuine minimal surfaces of the prescribed genus that the self-intersection numbers extracted from them are the same as those of the true surfaces.

What would settle it

A single knot for which the self-intersection count read off from the PINN surface differs from the integer predicted by the corresponding coefficient of the HOMFLY polynomial, or a rigorous bound showing that the PINN output deviates from the true minimal surface enough to change the intersection count.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • The PINN construction produces surfaces of the expected genus for every knot examined.
  • Signed self-intersection numbers can be extracted algorithmically from the numerical surfaces with sufficient reliability to match the conjecture.
  • Fine's conjecture receives empirical support across multiple knot families through exact numerical agreement.
  • The same computational pipeline supplies a practical method for testing the conjecture on further knots.

Where Pith is reading between the lines

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

  • If the approximation quality holds for larger or more complicated knots, the method could systematically test the conjecture where analytic constructions are unavailable.
  • The observed agreement indicates that the minimal-surface equation in H^4 encodes the same integer data carried by the HOMFLY polynomial.
  • The framework could be adapted to count minimal surfaces with other boundary conditions or in different ambient geometries that appear in related conjectures.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 2 minor

Summary. The paper introduces a Physics-Informed Neural Network (PINN) framework to approximate minimal surfaces in H^4 with prescribed genus that bound a given knot K at infinity in S^3. It develops an algorithmic procedure to extract signed self-intersection numbers from the learned parametrization and reports that, for every knot tested, these numbers exactly match the predictions of Fine's conjecture relating the HOMFLY polynomial coefficients to such counts, thereby supplying empirical evidence for the conjecture.

Significance. If the PINN approximations can be shown to be sufficiently accurate in a C^2 topology to guarantee that the extracted intersection counts are stable and correct, the work would constitute a novel computational bridge between knot invariants and the geometry of minimal surfaces in hyperbolic 4-space, offering a practical testbed for an otherwise analytically intractable conjecture.

major comments (3)
  1. [Abstract / numerical experiments section] Abstract and the section describing the numerical experiments: the headline claim that the discovered surfaces 'perfectly align' with Fine's conjecture requires that the learned surfaces are close enough to true minimal surfaces (in a norm controlling the zero set of the normal bundle or the algebraic intersection form) that small C^1 perturbations cannot create, destroy, or flip the sign of transverse intersections. No a-priori error bound, residual estimate in C^2, or convergence study is supplied to justify this stability.
  2. [Self-intersection extraction subsection] The subsection on the self-intersection extraction algorithm: the procedure for locating self-intersections and assigning signs from the neural-network parametrization is presented without a sensitivity analysis showing that the reported counts remain invariant under the observed residual of the minimal-surface PDE. Small discretization or approximation errors can alter transverse intersections, yet no such robustness check is performed.
  3. [Validation / experiments section] The validation paragraph (or its absence): the manuscript contains no comparison of the PINN solver against analytically known minimal surfaces in H^4 (e.g., geodesic spheres or Clifford tori) that would quantify how large the residual must be before the intersection count changes, leaving the reliability of the 'perfect alignment' claim unquantified.
minor comments (2)
  1. [Methods] Notation for the hyperbolic metric and the minimal-surface equation should be stated explicitly once at the beginning of the methods section rather than introduced piecemeal.
  2. [Numerical results] The description of the knot families tested would benefit from a table listing the specific knots, their HOMFLY coefficients, the predicted counts, and the numerically obtained counts side-by-side.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their careful reading and constructive comments, which highlight important aspects of numerical validation. We address each major comment below.

read point-by-point responses

  1. Referee: [Abstract / numerical experiments section] Abstract and the section describing the numerical experiments: the headline claim that the discovered surfaces 'perfectly align' with Fine's conjecture requires that the learned surfaces are close enough to true minimal surfaces (in a norm controlling the zero set of the normal bundle or the algebraic intersection form) that small C^1 perturbations cannot create, destroy, or flip the sign of transverse intersections. No a-priori error bound, residual estimate in C^2, or convergence study is supplied to justify this stability.

    Authors: We agree that the stability of the extracted intersection counts under the observed approximation error requires explicit support. In the revised manuscript we will add a convergence study reporting the PDE residual (in appropriate norms) together with the intersection counts as network depth and training iterations increase; we will also revise the abstract to qualify the alignment as holding for the computed surfaces with the residuals achieved. Deriving a priori C^2 error bounds for this PINN formulation lies outside the present scope. revision: partial

  2. Referee: [Self-intersection extraction subsection] The subsection on the self-intersection extraction algorithm: the procedure for locating self-intersections and assigning signs from the neural-network parametrization is presented without a sensitivity analysis showing that the reported counts remain invariant under the observed residual of the minimal-surface PDE. Small discretization or approximation errors can alter transverse intersections, yet no such robustness check is performed.

    Authors: We will add a sensitivity analysis in the revised version: small random perturbations to the network parameters, scaled to the magnitude of the observed residual, will be applied and the invariance of the signed intersection counts verified for the knots studied. revision: yes

  3. Referee: [Validation / experiments section] The validation paragraph (or its absence): the manuscript contains no comparison of the PINN solver against analytically known minimal surfaces in H^4 (e.g., geodesic spheres or Clifford tori) that would quantify how large the residual must be before the intersection count changes, leaving the reliability of the 'perfect alignment' claim unquantified.

    Authors: We will incorporate a validation subsection comparing the PINN solver against known minimal surfaces in H^4 (geodesic spheres and Clifford tori), measuring the residual at which the (known) intersection counts remain stable; this will supply a quantitative benchmark for the knot computations. revision: yes

Circularity Check

0 steps flagged

Empirical test of external conjecture; no internal derivation reduces to fitted inputs

full rationale

The paper introduces a PINN solver for the minimal surface equation in H^4 and uses it to generate numerical surfaces for specific knots, then extracts self-intersection counts and signs for comparison against Fine's Conjecture (an external statement by Joel Fine). No step claims to derive the conjecture coefficients or counts from quantities defined inside the paper; the central claim is an empirical match rather than a mathematical reduction. No self-citations, ansatzes, or fitted parameters are presented as load-bearing predictions. The work is therefore self-contained against external benchmarks and receives the default non-circularity finding.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities; the approach rests on the standard PINN framework and the external conjecture itself.

pith-pipeline@v0.9.1-grok · 5687 in / 1051 out tokens · 27190 ms · 2026-06-29T20:06:16.786566+00:00 · methodology

0 comments
read the original abstract

A recent conjecture by Joel Fine posits a relationship between the coefficients of the HOMFLY polynomial of a knot $K$ in the 3-sphere $S^3$, and the signed count of minimal surfaces in hyperbolic 4-space $\mathrm{H}^4$ meeting the sphere at infinity at $K$, with prescribed genus and self-intersection number. In this paper, we develop a novel machine learning framework based on Physics-Informed Neural Networks (PINNs) to solve the minimal surface equation in hyperbolic space. We utilise this framework to test Fine's Conjecture by constructing near-minimal surfaces bounding various families of knots in $S^3$. Furthermore, we develop an algorithmic method to find self-intersections and compute their sign. For every knot analysed, the computationally discovered minimal surfaces and their self-intersection numbers perfectly align with the predictions of Fine's Conjecture, providing empirical evidence for it.

Figures

Figures reproduced from arXiv: 2605.26234 by Marco Usula, Tancredi Schettini Gherardini.

Figure 2.1
Figure 2.1. Figure 2.1: An embedded circle, and an immersed figure-8. V ∈ C∞ (N; u ∗TM) as a vector field on M defined along the image u (N); for every point q ∈ u (N) of self-intersection, there is a vector for each “strand” of u (N) passing through q, according to the multiplicity of q. Given such a vector field V , one can displace u (N) along the unit-time flow of V and obtain a new smooth map u˜ : N → M. If V is small enou… view at source ↗
Figure 2.2
Figure 2.2. Figure 2.2: Knot diagrams of an unknot, a trefoil knot, and a Stevedore knot. cylinder ∂X ×[0, ε) embedded in a collar neighbourhood S n ×[0, ε) of S n in H n+1. If X◦ is minimal in H n+1, and ∂X is smoothly embedded in S n, then X is smooth in the interior, polyhomogeneous up to the boundary, and intersects the boundary at infinity orthogonally. Moreover, if m is even, then X is actually smooth up to the boundary. … view at source ↗
Figure 2.3
Figure 2.3. Figure 2.3: Different crossing types in the Skein relations. Definition 2.22. A link invariant is a map P : L → X from the space L of links in S 3 to a set X , such that if L1 ∼ L2 then P (L1) = P (L2). Similarly, an oriented link invariant is a map P : Lor → X from the space of oriented links Lor, such that if L1 or∼ L2 then P (L1) = P (L2) A way to prove that two links L1 and L2 are not isotopic is to exhibit a kn… view at source ↗
Figure 2.4
Figure 2.4. Figure 2.4: An immersed circle in R 2 with a triple point, resolved into three double points by a generic deformation. whose boundary map u∂ is an embedding, then the image u (Σ) is a p-immersed submanifold of H 4 intersecting the boundary at infinity at an oriented knot K = u∂ (∂Σ) ⊂ S 3 . Fine’s Conjecture proposes a relation between the coefficients of the HOMFLY polynomial of K, and an appropriate count of the n… view at source ↗
Figure 4.1
Figure 4.1. Figure 4.1: Four minimal discs in H 3 . 4. Results In this section, we discuss some of the solutions produced by our method. The model is, in principle, able to find p-immersed minimal discs in H n+1, for any n ≥ 2; for example, in [PITH_FULL_IMAGE:figures/full_fig_p026_4_1.png] view at source ↗
Figure 4.2
Figure 4.2. Figure 4.2: Knot diagrams of the knots tested in this paper. this problem as follows. Given a knot parametrisation γ : S 1 → R 3 , we deform it by adding a perturbation term σδ, with σ ∈ R + a scale parameter and δ : S 1 → R 3 a map of the form δ (θ) = X K n=1 An cos (nθ) + Bn sin (nθ). Here K ∈ N is fixed, and the Ai , Bi ∈ R 3 are randomly chosen vectors with entries in [0, 1]. For small enough values of σ, the ma… view at source ↗
Figure 4.3
Figure 4.3. Figure 4.3: |τ (uunknot)| 2 before (left) and after training (right). Note how the surface is minimal near the boundary even before training, by construction, as described in §3.2. A more quantitative assessment of the error after training is provided in [PITH_FULL_IMAGE:figures/full_fig_p028_4_3.png] view at source ↗
Figure 4.4
Figure 4.4. Figure 4.4: Monte Carlo simulation of L(uunknot; ·) with 1000 samples of size 2 14 . minimal discs. Our algorithm for the search of double points does not find any solution, confirming that the solution is indeed embedded. 4.2. Torus knots. Here we discuss solutions meeting the sphere at infinity along various torus knots. Definition 4.1. A torus knot is a knot that lies entirely on the surface of a torus in R 3 . A… view at source ↗
Figure 4
Figure 4. Figure 4: ), which all converge via Newton’s Method to a unique double point (shown in bright [PITH_FULL_IMAGE:figures/full_fig_p029_4.png] view at source ↗
Figure 4.5
Figure 4.5. Figure 4.5: Log self-proximity map of u3,2 and its double point. ≈ 10−16, that is 0 up to machine precision. We can then conclude that u3,2 is a near-minimal p-immersion D2 → H 4 with self-intersection number 1. The HOMFLY polynomial of 31 is P (31) = 2a 2 − a 4 + a 2 z 2 , and therefore the presence of the term 2a 2 predicts the existence of a minimal disc with self￾intersection number 1; our near-solution is consi… view at source ↗
Figure 4.6
Figure 4.6. Figure 4.6: Log self-proximity maps of u4,3, u˜4,3 and their double points. It should compared with [PITH_FULL_IMAGE:figures/full_fig_p031_4_6.png] view at source ↗
Figure 4.7
Figure 4.7. Figure 4.7: Log self-proximity maps of u5,2, u5,3 and their double points, match￾ing the prediction from Fine’s Conjecture. 4.3. Other knots. We discuss four more examples, namely the Figure-eight knot, the Three-twist knot, the Stevedore knot and the square knot, a non-prime knot with crossing number 6. None of these examples is a torus knot, and each of them explores different aspects of Fine’s Conjecture. The 41 … view at source ↗
Figure 4.8
Figure 4.8. Figure 4.8: Log self-proximity maps of ufigure8 (41), u3twist (52), uStevedore (61), usquare (31#3∗ 1 ) and their double points. Each result is predicted by Fine’s Con￾jecture. The 52 knot. The Three-twist knot is known as 52 in A–B notation. A diagram of this knot is shown in [PITH_FULL_IMAGE:figures/full_fig_p033_4_8.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. Black Hole Black Boxes: Numerical Black Hole Metrics via AInstein Neural Networks

    gr-qc 2026-07 conditional novelty 6.0

    Unsupervised Lorentzian PINNs with embedded S^{2} topology recover maximally extended Schwarzschild and yield candidate Petrov type-I vacuum black-hole metrics with genuinely trapped interiors.

Reference graph

Works this paper leans on

45 extracted references · 14 canonical work pages · cited by 1 Pith paper · 1 internal anchor

  1. [1]

    Machine learning sasakian andg 2 topology on contact calabi–yau 7-manifolds.arXiv preprint arXiv:2310.03064, 2023

    Arpit Aggarwal, Yang-Hui He, Jack Heyes, Edward Hirst, Henrique Sá Earp, and Andre Silva. Machine learning sasakian andg 2 topology on contact calabi–yau 7-manifolds.arXiv preprint arXiv:2310.03064, 2023

    work page arXiv 2023

  2. [2]

    J. W. Alexander. Topological invariants of knots and links.Trans. Am. Math. Soc., 30:275–306, 1928

    1928

  3. [3]

    High Energy Phys., 2021(5):45,

    LaraB.Anderson, MathisGerdes, JamesGray, SvenKrippendorf, NikhilRaghuram, andFabianRuehle.Moduli- dependent Calabi-Yau and SU(3)-structure metrics from machine learning.J. High Energy Phys., 2021(5):45,

    2021

  4. [4]

    Anderson

    Michael T. Anderson. Complete minimal varieties in hyperbolic space.Invent. Math., 69:477–494, 1982

    1982

  5. [5]

    Anthony Ashmore, Yang-Hui He, and Burt A. Ovrut. Machine Learning Calabi–Yau Metrics.Fortschritte der Physik, 68(9), 2020

    2020

  6. [6]

    Pearlmutter, Alexey Andreyevich Radul, and Jeffrey Mark Siskind

    Atılım Günes Baydin, Barak A. Pearlmutter, Alexey Andreyevich Radul, and Jeffrey Mark Siskind. Automatic differentiation in machine learning: a survey.Journal of Machine Learning Research, 18(153):1–43, 2018

    2018

  7. [7]

    Biquard, editor.AdS/CFT correspondence: Einstein metrics and their conformal boundaries

    O. Biquard, editor.AdS/CFT correspondence: Einstein metrics and their conformal boundaries. Proceedings, 73rd Meeting of Theoretical Physicists and Mathematicians, Strasbourg, France, September 11-13, 2003, Zurich, Switzerland, 2005. Eur. Math. Soc

    2003

  8. [8]

    Stapleton

    Gianfranco Cortes, Maria Esteban-Casadevall, Yueqing Feng, Jonas Henkel, Edward Hirst, Tancredi Schettini Gherardini, and Alexander G. Stapleton. A machine learning approach to the nirenberg problem.arXiv preprint arXiv:2602.12368, 2026

    work page arXiv 2026

  9. [9]

    Asymptotic plateau problem: a survey

    Baris Coskunuzer. Asymptotic plateau problem: a survey. InProceedings of the 20th Gökova geometry-topology conference, Gökova, Turkey, May 27 – June 1, 2013, pages 120–146. Somerville, MA: International Press; Gökova: Gökova Geometry-Topology Conferences (GGT), 2014

    2013

  10. [10]

    Aspects of holographic entanglement using physics-informed neural net- works.arXiv preprint arXiv:2509.25311, 2025

    Suvankar Deb and Rohan Sanghavi. Aspects of holographic entanglement using physics-informed neural net- works.arXiv preprint arXiv:2509.25311, 2025

    work page arXiv 2025

  11. [11]

    Douglas, Subramanian Lakshminarasimhan, and Yidi Qi

    Michael R. Douglas, Subramanian Lakshminarasimhan, and Yidi Qi. Numerical Calabi-Yau metrics from holo- morphic networks. InProceedings of the 2nd Mathematical and Scientific Machine Learning Conference, volume 145 ofProceedings of Machine Learning Research, pages 223–252, 2022

    2022

  12. [12]

    A physics-informed neural network framework for partial differential equations on 3d surfaces: Time-dependent problems.arXiv preprint arXiv:2103.13878, 2021

    Zhiwei Fang, Justin Zhang, and Xiu Yang. A physics-informed neural network framework for partial differential equations on 3d surfaces: Time-dependent problems.arXiv preprint arXiv:2103.13878, 2021

    work page arXiv 2021

  13. [13]

    Robin Graham

    Charles Fefferman and C. Robin Graham. Conformal invariants. Élie Cartan et les mathématiques d’aujourd’hui, The mathematical heritage of Élie Cartan, Sémin. Lyon 1984, Astérisque, No.Hors Sér. 1985, 95-116 (1985)., 1985

    1984

  14. [14]

    Robin Graham.The ambient metric, volume 178 ofAnnals of Mathematics Studies

    Charles Fefferman and C. Robin Graham.The ambient metric, volume 178 ofAnnals of Mathematics Studies. Princeton University Press, Princeton, NJ, 2012

    2012

  15. [15]

    Knots, minimal surfaces and J-holomorphic curves.arXiv preprint arXiv:2112.07713, 2021

    Joel Fine. Knots, minimal surfaces and J-holomorphic curves.arXiv preprint arXiv:2112.07713, 2021

    work page arXiv 2021

  16. [16]

    Freyd, D

    P. Freyd, D. Yetter, J. Hoste, W. B. R. Lickorish, K. Millett, and A. Ocneanu. A new polynomial invariant of knots and links.Bull. Am. Math. Soc., New Ser., 12:239–246, 1985

    1985

  17. [17]

    Computer-assisted proofs in pde: a survey.SeMA Journal, 76(3):459–484, 2019

    Javier Gomez-Serrano. Computer-assisted proofs in pde: a survey.SeMA Journal, 76(3):459–484, 2019

    2019

  18. [18]

    Robin Graham and John M

    C. Robin Graham and John M. Lee. Einstein metrics with prescribed conformal infinity on the ball.Adv. Math., 87(2):186–225, 1991

    1991

  19. [19]

    SIAM, Philadelphia, PA, 2nd edition, 2008

    Andreas Griewank and Andrea Walther.Evaluating Derivatives: Principles and Techniques of Algorithmic Differentiation. SIAM, Philadelphia, PA, 2nd edition, 2008

    2008

  20. [20]

    Metric flows with neural networks.arXiv preprint arXiv:2310.19870, 2023

    James Halverson and Fabian Ruehle. Metric flows with neural networks.arXiv preprint arXiv:2310.19870, 2023

    work page arXiv 2023

  21. [21]

    Approximating Continuous Functions by ReLU Nets of Minimal Width, 2018

    Boris Hanin and Mark Sellke. Approximating Continuous Functions by ReLU Nets of Minimal Width, 2018

    2018

  22. [22]

    MINIMAL SURFACES, KNOTS, AND NEURAL NETWORKS 37

    KojiHashimoto, KoichiKyo, MasakiMurata, GakutoOgiwara, andNorihiroTanahashi.Physics-informedneural network solves minimal surfaces in curved spacetime.arXiv preprint arXiv:2509.10866, 2025. MINIMAL SURFACES, KNOTS, AND NEURAL NETWORKS 37

    work page arXiv 2025

  23. [23]

    Minimising Willmore Energy via Neural Flow

    Edward Hirst, Henrique N. Sá Earp, and Tomás S. R. Silva. Minimising willmore energy via neural flow.arXiv preprint arXiv:2604.04321, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  24. [24]

    Stapleton

    Edward Hirst, Tancredi Schettini Gherardini, and Alexander G. Stapleton. Ainstein: Numerical einstein metrics via machine learning.arXiv preprint arXiv:2502.13043, 2025

    work page arXiv 2025

  25. [25]

    Multilayer feedforward networks are universal approx- imators.Neural Networks, 2(5):359–366, 1989

    Kurt Hornik, Maxwell Stinchcombe, and Halbert White. Multilayer feedforward networks are universal approx- imators.Neural Networks, 2(5):359–366, 1989

    1989

  26. [26]

    Neural network approximations for Calabi-Yau metrics.arXiv, 08:105, 2022

    Vishnu Jejjala, Damian Kaloni Mayorga Pena, and Challenger Mishra. Neural network approximations for Calabi-Yau metrics.arXiv, 08:105, 2022

    2022

  27. [27]

    V. F. R. Jones. Hecke algebra representations of braid groups and link polynomials.Ann. Math. (2), 126:335–388, 1987

    1987

  28. [28]

    Vaughan F. R. Jones. A polynomial invariant for knots via von Neumann algebras.Bull. Am. Math. Soc., New Ser., 12:103–111, 1985

    1985

  29. [29]

    Universal approximation with deep narrow networks, 2020

    Patrick Kidger and Terry Lyons. Universal approximation with deep narrow networks, 2020

    2020

  30. [30]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization.Proceedings of ICLR, 2015

    2015

  31. [31]

    Learning Size and Shape of Calabi-Yau Spaces,

    Magdalena Larfors, Andre Lukas, Fabian Ruehle, and Robin Schneider. Learning size and shape of calabi-yau spaces.arXiv preprint arXiv:2111.01436, 2021

    work page arXiv 2021

  32. [32]

    DongC.LiuandJorgeNocedal.OnthelimitedmemoryBFGSmethodforlargescaleoptimization.Mathematical Programming, 45(3):503–528, 1989

    1989

  33. [33]

    Bruno De Luca

    G. Bruno De Luca. Machine learning gravity compactifications on negatively curved manifolds.arXiv preprint arXiv:2501.00093, 2024

    work page arXiv 2024

  34. [34]

    Variations of renormalized volume for minimal submanifolds of Poincaré-Einstein manifolds

    Jared Marx-Kuo. Variations of renormalized volume for minimal submanifolds of Poincaré-Einstein manifolds. Commun. Anal. Geom., 33(1):17–29, 2025

    2025

  35. [35]

    Mazzeo.Hodge cohomology of negatively curved manifolds

    Rafe R. Mazzeo.Hodge cohomology of negatively curved manifolds. ProQuest LLC, Ann Arbor, MI, 1986. Thesis (Ph.D.)–Massachusetts Institute of Technology

    1986

  36. [36]

    Mazzeo and Richard B

    Rafe R. Mazzeo and Richard B. Melrose. Meromorphic extension of the resolvent on complete spaces with asymptotically constant negative curvature.J. Funct. Anal., 75(2):260–310, 1987

    1987

  37. [37]

    Richard B. Melrose. Differential analysis on manifolds with corners, 1996

    1996

  38. [38]

    Updating quasi-Newton matrices with limited storage.Mathematics of Computation, 35(151):773–782, 1980

    Jorge Nocedal. Updating quasi-Newton matrices with limited storage.Mathematics of Computation, 35(151):773–782, 1980

    1980

  39. [39]

    Non-uniqueness and symmetries for the nirenberg problem using computer assistance.arXiv pre- print arXiv:2603.29544, 2026

    Daniel Platt. Non-uniqueness and symmetries for the nirenberg problem using computer assistance.arXiv pre- print arXiv:2603.29544, 2026

    work page arXiv 2026

  40. [40]

    Raissi, P

    M. Raissi, P. Perdikaris, and G. E. Karniadakis. Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations.J. Comput. Phys., 378:686–707, 2019

    2019

  41. [41]

    Biharmonic maps between conformally compact manifolds.Calc

    Marco Usula. Biharmonic maps between conformally compact manifolds.Calc. Var. Partial Differ. Equ., 65(2):24, 2026. Id/No 59

    2026

  42. [42]

    Isometric embeddings of conformally compact manifolds into hyperbolic spaces.International Mathematics Research Notices, 2026(10):rnag098, 05 2026

    Marco Usula. Isometric embeddings of conformally compact manifolds into hyperbolic spaces.International Mathematics Research Notices, 2026(10):rnag098, 05 2026

    2026

  43. [43]

    Yongji Wang, Mehdi Bennani, James Martens, Sébastien Racanière, Sam Blackwell, Alex Matthews, Stan- islav Nikolov, Gonzalo Cao-Labora, Daniel S. Park, Martin Arjovsky, Daniel Worrall, Chongli Qin, Ferran Alet, Borislav Kozlovskii, Nenad Tomašev, Alex Davies, Pushmeet Kohli, Tristan Buckmaster, Bogdan Geor- giev, Javier Gómez-Serrano, Ray Jiang, and Ching-...

    work page arXiv 2025

  44. [44]

    Convergence conditions for ascent methods.SIAM Review, 11(2):226–235, 1969

    Philip Wolfe. Convergence conditions for ascent methods.SIAM Review, 11(2):226–235, 1969

    1969

  45. [45]

    Approximating high-dimensional minimal surfaces with physics-informed neural networks.arXiv preprint arXiv:2309.02589, 2023

    Steven Zhou and Xiaojing Ye. Approximating high-dimensional minimal surfaces with physics-informed neural networks.arXiv preprint arXiv:2309.02589, 2023

    work page arXiv 2023