Pith. sign in

REVIEW 4 major objections 5 minor 48 references

Limits of the inverse scattering problem

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read The paper claims that finite-speed Newtonian scattering can recover a force field exactly when the particle's kinetic energy exceeds the potential gap between the interior and the boundary of the region.

desk verdict Solid examples, shaky universal threshold: the harmonic and constant-force calculations are real contributions, but Eq. (1.1) is a plausible heuristic, not an established limit. read the letter →

arxiv 2608.13105 v1 pith:YDRQSAWA submitted 2026-08-13 hep-th math-phmath.MP

classification hep-thmath-phmath.MP MSC 35R3044A1281U4065R32
keywords inversescatteringNewtoniantomographyvectorrayintegralRadontransformfinite-speedparticlespotentialreconstructioncriticalenergythresholdmachinelearning
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

The paper studies a Newtonian version of tomography: particles are fired into a disk and the unknown conservative force field inside is reconstructed from boundary data such as entry point, initial speed, traversal time, exit point, and exit velocity. The authors claim that this reconstruction is possible only when the particle's kinetic energy exceeds a characteristic potential gap between the interior and the boundary, $E=\max(|U_{\rm in}-U_{\rm out}|)$. Two exactly solvable examples, a harmonic well and a constant force, together with a machine-learning reconstruction pipeline, support the claim: above the threshold the inverse problem behaves like a corrected Radon transform, while below it the scattering map loses coverage and the reconstruction error rises steeply. If the claim is right, it gives a practical lower bound on beam energy for particle-based tomography rather than an absolute speed scale.

What carries the argument

The load-bearing object is the vector ray integral $J_v(L)=v\,\Delta v(L)$, the speed-weighted momentum change of a particle crossing the disk along a chord $L$; in the infinite-speed limit it reduces to the Radon transform of the force field, which is inverted componentwise. The argument proceeds by a fixed-boundary perturbative expansion in $\varepsilon=\delta U/(mv^2)$: the trajectory and flight time are expanded with pinned entry and exit points, generating explicit functionals $Q[F]$ and $P[F]$ that correct the ray integral at finite speed. The harmonic and constant-force problems are solved exactly and confirm the expansion; the machine-learning pipeline, the Force-Field Prediction Model, reconstructs the force field from boundary measurements plus a Radon prior, with the kinetic-to-potential energy ratio $\kappa=E_{\rm kin}/E_{\rm pot}$ as the controlling parameter.

What would settle it

Construct a disk potential containing a deep, wide well surrounded by a high but narrow barrier, and launch particles with energy above the local potential difference inside the well; if angular momentum prevents them from crossing the barrier, the reconstruction should fail despite satisfying Eq. (1.1). A numerical experiment of this kind would settle whether the energy criterion is sufficient or only necessary.

Watch

Extended reading notes

Core claim

The central claim is that finite-speed Newtonian boundary scattering can recover a conservative force field on the unit disk exactly when the initial kinetic energy is at least the absolute potential variation between the interior and the boundary, as stated in Eq. (1.1). The paper develops a systematic expansion of the measured vector ray integral, $J_v = v\,\Delta v = J_\infty + (\delta U/mv^2)Q[F] + (\delta U/mv^2)^2P[F] + \cdots$, about the infinite-speed Radon limit, with the correction functionals $Q[F]$ and $P[F]$ integrated along the unperturbed chord. For the harmonic potential $U=r^2$, the first correction vanishes, the next term is $J_\infty^3/(8v^4)$, and the series converges only for $v\ge\sqrt{2}$ when $\delta U=m=1$; this same speed separates two-to-one scattering coverage from one-to-one coverage. Numerical experiments on synthetic random fields and on a held-out paraboloid show reconstruction error growing sharply in the window $E_{\rm kin}/E_{\rm pot}\in[1,3]$, which brackets the predicted threshold.

Load-bearing premise

The threshold assumes that a particle with kinetic energy above a local potential difference can actually reach every interior point, ignoring angular-momentum barriers and topological obstructions that can keep particles out of some regions; the paper states these restrictions are intentionally ignored.

Editorial extensions

If this is right

  • Above the threshold, the inverse Radon formula used for X-rays can be corrected order by order in $1/v^2$, giving an iterative reconstruction scheme that converges for speeds above a critical value.
  • Below the threshold, some exit points are unreachable or multiply covered, so the boundary data no longer determine the force field; for the harmonic well the dividing speed is exactly $v_c=\sqrt{2\,\delta U/m}$.
  • Measurements at two different speeds can be combined algebraically to eliminate the first correction and recover the field directly, without knowing the correction functional in advance.
  • Reconstruction quality is governed by the ratio $E_{\rm kin}/E_{\rm pot}$, so the same transition is predicted for slow particles in shallow potentials and fast particles in deep ones.
  • The paper conjectures the universal nonlinear relation $J_\infty = \frac{2(1-|J_v|^2/(4v^4))}{\sqrt{4-(J_v\cdot n)^2/v^4}-J_v\cdot\tau/v^2}\,J_v$, and shows it is exact for the harmonic and constant-force cases, suggesting a data-only correction step before Radon inversion.

Reading between the lines

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

  • A testable extension: in non-central potentials the local energy indicator ignores angular-momentum barriers, so the true recoverability threshold may exceed Eq. (1.1); a simulation with a deep well shielded by a narrow barrier would separate the energy condition from the reachability condition.
  • The residence-time channel is the boundary measurement most correlated with field amplitude, so time-of-flight data alone may serve as a practical detector of the threshold in experimental setups.
  • If the conjectured nonlinear relation between $J_v$ and $J_\infty$ holds approximately for general fields, it would provide a preprocessing step that removes most finite-speed distortion before any Radon inversion.
  • A practical design rule follows: choose the beam energy to exceed the maximum interior-to-boundary potential drop of the target potential class, rather than tuning to an absolute speed.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper studies the inverse scattering problem for Newtonian particles at finite speed on the unit disk, asking how low the initial kinetic energy can be while the conservative force field remains reconstructible from boundary-to-boundary flight data. After recalling the infinite-speed Radon limit, the authors develop a perturbative expansion of the finite-speed vector ray integral in powers of δU/(mv²), analyze two exactly solvable examples (harmonic potential and constant force), conjecture a universal non-linear relation between finite- and infinite-speed ray integrals (Eq. (6.1)), and train a GAN-based Force-Field Prediction Model on synthetic random potentials. The central claim is Eq. (1.1): the minimal kinetic energy required to restore the potential equals max(|U_in − U_out|). The harmonic example is treated in detail and reproduces the critical speed v_c = √2; the constant-force example, the ML experiments, and the out-of-distribution paraboloid test are used as additional evidence for the same energy-scale transition.

Significance. If Eq. (1.1) were established, it would give a simple and practically attractive criterion for planning finite-speed particle-tomography experiments: the required kinetic energy is set by the maximal potential difference between boundary and interior. The paper has real strengths: the harmonic-oscillator analysis in Section 4 is explicit and internally consistent; the perturbative framework suggests useful single-speed iteration and multi-speed extrapolation procedures; and the authors provide code and datasets, with a substantial ML engineering effort. However, as written, Eq. (1.1) is a conjecture rather than a proven limit. One of the two exact examples contains a concrete algebraic error, and the conjectured universal formula of Section 6 is contradicted by the corrected constant-force kinematics. The ML validation is also partly circular. The paper is best viewed as a well-motivated conjecture with supporting numerics, not an established theorem.

major comments (4)
  1. [§5.1, Eq. (5.8)] The exact constant-force ray integral is dimensionally inconsistent and does not reduce to the infinite-speed limit. With m=1, J_v has units of velocity squared, but the right-hand side of Eq. (5.8), √2 v²/g times a quantity of units velocity, has units length×velocity and diverges as Lv as v→∞, instead of approaching J∞=Lg of Eq. (5.9). The correct expression is J_v(L)=√2 v √(v²+Lg∥−√((v²+Lg∥)²−g²L²)) (g/g). Expanding the corrected expression gives J_v = Lg − L² g∥ g/(2v²)+O(v^{-4}), with the opposite sign to Eq. (5.10). Consequently the claimed agreement of Q[F] with the general formula in §5.2, and the support this example is said to lend to Eq. (1.1), are not valid as written.
  2. [§6, Eq. (6.1)] The conjectured universal relation (6.1) is not exact for the constant-force family, contrary to the claim that algebraic inversion of (5.8) leads to it. For a constant force parallel to the chord, the exact kinematics give J∞ = J_v + J_v²/(2v²) (as vector relation along τ), whereas Eq. (6.1) with J_v·n=0 reduces to J_v√(1−J_v²/(4v⁴)). The two differ already at O(v^{-2}). Thus Eq. (6.1) is not a general exact relation even for the two solvable families; it is exact only for the harmonic case, where it collapses to Eq. (6.2). This undermines the paper's presentation of (6.1) as a universal structure.
  3. [Eq. (1.1) and §7.7.1] The central claim Eq. (1.1) is not established for general potentials. The perturbative expansion of §3.2 has no convergence theorem for general fields; the only critical-speed derivation is the harmonic example of Section 4. Section 7.7.1 explicitly states that the local energy indicator f_acc 'does not take into account the connectivity of the set of reachable points' and that topological restrictions are 'intentionally ignored.' Angular-momentum barriers can therefore exclude interior regions even when the pointwise energy inequality holds, so sufficiency of Eq. (1.1) is assumed rather than proven. The numerical support is partly circular: the constants C_1..C_5 in Eq. (7.12) are computed from the same synthetic-field ensemble used for validation, metric 4 is selected post hoc, and Figs. 5–6 min-max scale the 'inaccessible area' curve to fit the error curve, so no independent falsifiable test of Eq. (1.1) is provided. The paper should either prove the criterion for a well-defined class (for example monotone central potentials), or state Eq. (1.1) as a conjecture and validate it with an independently calibrated, pre-registered energy scale.
  4. [§3.2, Eqs. (3.12)–(3.29)] The perturbative derivation, while suggestive, is not self-contained enough to justify the transition claim. The half-integer expansion of T is justified by a described iterative procedure but not written out, no general convergence radius ϵ_c is given for arbitrary F, and the value ϵ_c=1/2 is obtained only for the harmonic oscillator. The sentence 'Note that T^2 = ...' appears in the middle of the derivation without a clear role and reads as an unfinished note. Since Eq. (3.29) is the theoretical bridge to Eq. (1.1), this gap should be addressed directly rather than left as an implicit assumption.
minor comments (5)
  1. [§1, author line] The author names and affiliations contain corrupted characters ('Fediniπe', 'iInstitute') that should be cleaned before publication.
  2. [§7.6, Fig. 4] The Img2Vec amplitude prediction is shown only as a scatter plot; please report quantitative metrics such as R² or mean absolute error in log10A.
  3. [§7.7, Figs. 5 and 6] The 'inaccessible area, min–max scaled to fit' curves are rescaled to match the error curves; this makes the visual agreement uninformative. The inaccessible fraction should be shown on its own axis or with a stated, physically motivated normalization.
  4. [§8, Conclusions] The statement that future work will address 'independent test sets and repeated training runs' confirms that the current ML results come from single training runs; this limitation should be stated in Section 7.7.1 rather than only in the conclusions.
  5. [§3.2, Eq. (3.19)] The notation uses v both as the scalar launch speed and implicitly inside the vector v_in; please distinguish the scalar |v| from the vector v_in to avoid confusion.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Eq. (1.1) is a conjecture supported by exact analytic examples and an ML characterization; the same-data energy calibration and post-hoc metric choice weaken the empirical support but do not make the derivation circular.

full rationale

The central claim, Eq. (1.1), is not derived from its own validation in a circular way. In the harmonic example, the paper derives an exact critical velocity v_c = sqrt(2) from both the convergence of the 1/v series (Section 4.3, Eq. (4.21) and the following text) and from the change in the topological coverage of the scattering map (Section 4.1). The energy threshold E = U(1)-U(0) = 1 then follows from these independent exact calculations, and the constant-force example gives a matching chord-dependent critical speed. Neither example defines E in terms of the quantity being predicted. The ML section uses energy scales E_pot^(k) = C_k A (Eq. (7.12)) with constants evaluated from the same synthetic fields, but those constants are statistical normalizations of the x-axis, not fitted parameters of a prediction; the reconstruction error itself is measured independently. The subsequent choice of metric 4 as the preferred definition is a post-hoc selection that weakens the empirical evidence, but it does not make Eq. (1.1) equivalent to the input by construction. Section 7.7.1 explicitly concedes that the local energy indicator 'does not take into account the connectivity of the set of reachable points' and that topological restrictions are 'intentionally ignored'; this is a genuine limitation on the universality and sufficiency of Eq. (1.1), but it is a scope restriction rather than a circular step. The paper also contains no load-bearing self-citations: the external uniqueness theorems it invokes (Novikov, Jollivet, Pestov-Uhlmann, etc.) are independent mathematical results, and the only self-citation is [17], a code and data repository. Overall, the derivation chain is self-contained for the two solvable examples; the generalization to a universal threshold is an under-supported conjecture, but under-support is a correctness risk, not circularity.

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

The theoretical derivation uses standard Newtonian mechanics and Radon inversion. The main extra assumptions are the sufficiency of the energy criterion and the interpretation of perturbative convergence as the recoverability boundary. The ML validation introduces empirical energy-scale constants calibrated on the evaluation dataset.

free parameters (2)
  • characteristic potential variation delta_U = taken as 1 in the harmonic example; general case unspecified
    Used to define epsilon = delta_U/(m v^2). The central threshold E = max(|U_in - U_out|) depends on choosing this scale, which is ambiguous for general potentials.
  • ensemble energy-scale constants C1..C5 = C1=0.551, C2=0.973, C3=0.272, C4=0.497, C5=0.225
    Computed from the normalized synthetic field ensemble and used to set kappa = E_kin/E_pot in validation. They are calibrated to the data, and metric 4 is selected post hoc.
assumptions (4)
  • domain assumption Newtonian equations of motion with conservative force F = -grad U on the unit disk
    Section 3, Eq. (3.1).
  • domain assumption Boundary scattering data (time, exit angle, exit velocity) over all launch directions determine the vector ray integral J_v
    Section 3, list of measurable properties.
  • ad hoc to paper Perturbative expansion J_v = J_inf + epsilon Q + epsilon^2 P converges for v above a critical speed, and failure of convergence marks loss of recoverability
    Sections 3.2 and 4.3. The bridge from series convergence to recoverability is asserted, not proven.
  • ad hoc to paper Local energy condition E_kin >= Delta U(x) is sufficient for a point to be probed; angular-momentum and topological barriers are ignored
    Section 7.7.1 explicitly states that topological restrictions are intentionally ignored; this is load-bearing for Eq. (1.1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Limits of the inverse scattering problem." pith.science (2026). https://pith.science/paper/YDRQSAWA

@misc{pith2026260813105,
  author       = {Pith},
  title        = {Pith review of: Limits of the inverse scattering problem},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YDRQSAWA}},
  note         = {Machine review of arXiv:2608.13105}
}
read the original abstract

The main goal of tomography is the reconstruction of density function out of its line integrals (integral measurements of this density along x-ray lines). Such construction is possible and known as inverse x-ray/Radon transformations. We are interested in generalization of this problem to the case of particles. For this we need to limit the speed of these particles, otherwise the problem is reduced to the previous one. The question we discuss in this paper is how low can this speed be depending on the parameters of the studied potential. We study this problem using simple theoretical examples and Machine Learning pipeline to restore the potential.

Figures

Figures reproduced from arXiv: 2608.13105 by the authors.

Figure 1
Figure 1. Angular parametrization of the particle launch geometry. The angle [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Selecting a working configuration: (a) time-to-quality comparison of residual FFPM [PITH_FULL_IMAGE:figures/full_fig_p024_2.png] view at source ↗
Figure 3
Figure 3. FFPM training and evaluation pipeline. As shown in [PITH_FULL_IMAGE:figures/full_fig_p025_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Prediction of log10 A by the auxiliary Img2Vec regressor from the measurement tensor D, without using the Radon reconstruction. The dashed diagonal denotes exact prediction. The auxiliary Img2Vec model was trained on a separate dataset, covering the range A ∈ [10−12 , …
Figure 5
Figure 5. Figure 5: Potential-reconstruction error as a function of [PITH_FULL_IMAGE:figures/full_fig_p029_5.png]
Figure 6
Figure 6. Figure 6: Smoothed reconstruction-error dependence for the selected interior-to-boundary [PITH_FULL_IMAGE:figures/full_fig_p029_6.png]
Figure 7
Figure 7. Figure 7: Out-of-distribution reconstruction of a centered two-dimensional paraboloid. [PITH_FULL_IMAGE:figures/full_fig_p030_7.png]
Figure 8
Figure 8. Figure 8: True, Radon-reconstructed, and FFPM-reconstructed centered paraboloid potentials [PITH_FULL_IMAGE:figures/full_fig_p031_8.png]
Figure 9
Figure 9. Figure 9: Detailed FFPM architecture. The generator follows an encoder–residual-bottleneck– [PITH_FULL_IMAGE:figures/full_fig_p034_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 33 canonical work pages

  1. [1]

    Colton and R

    D. Colton and R. Kress. Inverse Scattering. In O. Scherzer, editor,Handbook of Math- ematical Methods in Imaging, 2nd edition, pages 649–700. Springer, 2015.https: //doi.org/10.1007/978-1-4939-0790-8_48

  2. [2]

    Faridani

    A. Faridani. Introduction to the Mathematics of Computed Tomography.https://sites. science.oregonstate.edu/˜faridani/preprints/AFintroct.pdf

  3. [3]

    Jollivet

    A. Jollivet. On Inverse Scattering for the Multidimensional Relativistic Newton Equation at High Energies.Journal of Mathematical Physics, 47(6), 2006.https://doi.org/10. 1063/1.2206690

  4. [4]

    J. B. Keller, I. Kay, and J. Shmoys. Determination of the Potential from Scattering Data. Physical Review, 102:557–559, 1956.https://doi.org/10.1103/PhysRev.102.557

  5. [5]

    H. E. Moses. Calculation of the Scattering Potential from Reflection Coefficients.Physical Review, 102:559–567, 1956.https://doi.org/10.1103/PhysRev.102.559

  6. [6]

    R. G. Novikov. Small Angle Scattering and X-Ray Transform in Classical Mechanics.Arkiv for Matematik, 37(1):141–169, 1999.https://doi.org/10.1007/BF02384831

  7. [7]

    On inverse problems in electromagnetic field in classical mechanics at fixed energy

    A. Jollivet. On Inverse Problems in Electromagnetic Field in Classical Mechanics at Fixed Energy. 2007.https://arxiv.org/abs/math-ph/0701008

  8. [8]

    Kress and W

    R. Kress and W. Rundell. Inverse Obstacle Scattering Using Reduced Data.SIAM Journal on Applied Mathematics, 59(2):442–454, 1998.https://www.jstor.org/stable/118431

Show all 48 references
  1. [9]

    Pestov and G

    L. Pestov and G. Uhlmann. Two-Dimensional Compact Simple Riemannian Manifolds Are Boundary Distance Rigid. 2003.https://arxiv.org/abs/math/0305280

  2. [10]

    Stefanov and G

    P. Stefanov and G. Uhlmann. Boundary Rigidity and Stability for Generic Simple Metrics. 2004.https://arxiv.org/abs/math/0408075

  3. [11]

    R. G. Muhometov. The Problem of Reconstructing a Two-Dimensional Riemannian Met- ric and Integral Geometry.Doklady Akademii Nauk SSSR, 232(1):32–35, 1977. English translation:Soviet Mathematics Doklady, 18(1):27–31, 1977.http://mi.mathnet.ru/ dan40848

  4. [12]

    Adler and O

    J. Adler and O. Oktem. Learned Primal-Dual Reconstruction.IEEE Transactions on Medical Imaging, 37(6):1322–1332, 2018.https://doi.org/10.1109/TMI.2018.2799231 36

  5. [13]

    Tsang, O

    O. Tsang, O. Melia, V. Charisopoulos, J. Hoskins, Y. Khoo, and R. Willett. A Model- Guided Neural Network Method for the Inverse Scattering Problem. 2025.https:// arxiv.org/abs/2512.10123

  6. [14]

    Khachi and G

    A. Khachi and G. Balassa. Estimating Inverse Scattering Potentials for then-pSystem Using Variational Monte Carlo Neural Networks. 2024.https://arxiv.org/abs/2407. 02137

  7. [15]

    Lantz, N

    V. Lantz, N. Abiri, G. Carlsson, and M.-E. Pistol. Deep Learning for Inverse Problems in Quantum Mechanics.International Journal of Quantum Chemistry, 121(9):e26599, 2021. https://doi.org/10.1002/qua.26599

  8. [16]

    Jeong, H

    H.-S. Jeong, H. Kim, K.-Y. Kim, G. Yun, H. Yu, and K. Yun. AdS/Deep-Learning Made Easy II: Neural Network-Based Approaches to Holography and Inverse Problems. 2025. https://arxiv.org/abs/2511.22522

  9. [17]

    Fedin, K

    M. Fedin, K. Gubarev,A. Morozov. Limits of the inverse scattering problem.Zenodo, 2026. https://doi.org/10.5281/zenodo.21909377

  10. [18]

    van Aarle, W

    W. van Aarle, W. J. Palenstijn, J. Cant, E. Janssens, F. Bleichrodt, A. Dabravolski, J. De Beenhouwer, K. J. Batenburg, and J. Sijbers. Fast and Flexible X-Ray Tomography Using the ASTRA Toolbox.Optics Express, 24(22):25129–25147, 2016.https://doi.org/10. 1364/OE.24.025129

  11. [19]

    van der Walt, J

    S. van der Walt, J. L. Sch¨ onberger, J. Nunez-Iglesias, F. Boulogne, J. D. Warner, N. Yager, E. Gouillart, T. Yu, and the scikit-image contributors. scikit-image: Image Processing in Python.PeerJ, 2:e453, 2014.https://doi.org/10.7717/peerj.453

  12. [20]

    Johnson, A

    J. Johnson, A. Alahi, and L. Fei-Fei. Perceptual Losses for Real-Time Style Transfer and Super-Resolution. InEuropean Conference on Computer Vision, pages 694–711, 2016. https://doi.org/10.1007/978-3-319-46475-6_43

  13. [21]

    Isola, J.-Y

    P. Isola, J.-Y. Zhu, T. Zhou, and A. A. Efros. Image-to-Image Trans- lation with Conditional Adversarial Networks. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 1125– 1134, 2017.https://openaccess.thecvf.com/content_cvpr_2017/htm...

  14. [22]

    Wang, M.-Y

    T.-C. Wang, M.-Y. Liu, J.-Y. Zhu, A. Tao, J. Kautz, and B. Catanzaro. High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 37 8798–8807, 2018.https://opena...

  15. [23]

    Ronneberger, P

    O. Ronneberger, P. Fischer, and T. Brox. U-Net: Convolutional Networks for Biomedical Image Segmentation. InMedical Image Computing and Computer-Assisted Intervention, pages 234–241, 2015.https://doi.org/10.1007/978-3-319-24574-4_28

  16. [24]

    Peebles and S

    W. Peebles and S. Xie. Scalable Diffusion Models with Transformers. InProceed- ings of the IEEE/CVF International Conference on Computer Vision, pages 4195–4205, 2023.https://openaccess.thecvf.com/content/ICCV2023/html/Peebles_Scalable_ Diffusion_Models_with_Transformers_ICCV_...

  17. [25]

    van den Oord, Y

    A. van den Oord, Y. Li, and O. Vinyals. Representation Learning with Contrastive Pre- dictive Coding.arXiv preprint arXiv:1807.03748, 2018.https://arxiv.org/abs/1807. 03748

  18. [26]

    T. Park, A. A. Efros, R. Zhang, and J.-Y. Zhu. Contrastive Learning for Unpaired Image- to-Image Translation. InEuropean Conference on Computer Vision, pages 319–345, 2020. https://arxiv.org/abs/2007.15651

  19. [27]

    Edelsbrunner, D

    H. Edelsbrunner, D. Letscher, and A. Zomorodian. Topological Persistence and Simplifi- cation.Discrete & Computational Geometry, 28(4):511–533, 2002.https://doi.org/10. 1007/s00454-002-2885-2

  20. [28]

    Cohen-Steiner, H

    D. Cohen-Steiner, H. Edelsbrunner, and J. Harer. Stability of Persistence Diagrams. Discrete & Computational Geometry, 37(1):103–120, 2007.https://doi.org/10.1007/ s00454-006-1276-5

  21. [29]

    C. E. Rasmussen and C. K. I. Williams.Gaussian Processes for Machine Learning. MIT Press, 2006.https://gaussianprocess.org/gpml/chapters/

  22. [30]

    Kanagawa, P

    M. Kanagawa, P. Hennig, D. Sejdinovic, and B. K. Sriperumbudur. Gaussian Processes and Kernel Methods: A Review on Connections and Equivalences. 2018.https://arxiv. org/abs/1807.02582

  23. [31]

    Guttorp and T

    P. Guttorp and T. Gneiting. Studies in the History of Probability and Statistics XLIX: On the Matern Correlation Family.Biometrika, 93(4):989–995, 2006.https://doi.org/ 10.1093/biomet/93.4.989

  24. [32]

    Lindgren, H

    F. Lindgren, H. Rue, and J. Lindstrom. An Explicit Link Between Gaussian Fields and Gaussian Markov Random Fields: The Stochastic Partial Differential Equation Approach. Journal of the Royal Statistical Society: Series B, 73(4):423–498, 2011.https://doi. org/10.1111/j.1467-986...

  25. [33]

    Lindgren, D

    F. Lindgren, D. Bolin, and H. Rue. The SPDE Approach for Gaussian and Non-Gaussian Fields: 10 Years and Still Running.Spatial Statistics, 50:100599, 2022.https://arxiv. org/abs/2111.01084

  26. [34]

    Shinozuka and G

    M. Shinozuka and G. Deodatis. Simulation of Multi-Dimensional Gaussian Stochastic Fields by Spectral Representation.Applied Mechanics Reviews, 49(1):29–53, 1996.https: //doi.org/10.1115/1.3101883

  27. [35]

    Kurbanmuradov, K

    O. Kurbanmuradov, K. Sabelfeld, and P. R. Kramer. Randomized Spectral and Fourier- Wavelet Methods for Multidimensional Gaussian Random Vector Fields.Journal of Com- putational Physics, 2013.https://doi.org/10.1016/j.jcp.2013.03.021

  28. [36]

    Chepurnov

    A. Chepurnov. Explicit Form of the Random Field Spectral Representation and Some Applications. 2021.https://arxiv.org/abs/2107.05070

  29. [37]

    P. J. Green. Introduction to Finite Mixtures. 2018.https://arxiv.org/abs/1705.01505

  30. [38]

    G. J. McLachlan, S. X. Lee, and S. I. Rathnayake. Finite Mixture Models.Annual Review of Statistics and Its Application, 6:355–378, 2019.https://doi.org/10.1146/ annurev-statistics-031017-100325

  31. [39]

    K. Perlin. An Image Synthesizer.ACM SIGGRAPH Computer Graphics, 19(3):287–296, 1985.https://www.cs.cmu.edu/afs/cs/academic/class/15869-f11/www/readings/ perlin85_imagesynthesizer.pdf

  32. [40]

    K. Perlin. Improving Noise.ACM Transactions on Graphics, 21(3):681–682, 2002.https: //doi.org/10.1145/566654.566636

  33. [41]

    Lagae, S

    A. Lagae, S. Lefebvre, R. Cook, T. DeRose, G. Drettakis, D. S. Ebert, J. P. Lewis, K. Per- lin, and M. Zwicker. A Survey of Procedural Noise Functions.Computer Graphics Forum, 29(8), 2010.https://doi.org/10.1111/j.1467-8659.2010.01827.x

  34. [42]

    A. Jain, A. Sharma, and K. S. Rajan. Adaptive and Multi-Resolution Procedural Infinite Terrain Generation with Diffusion Models and Perlin Noise. InProceedings of the Thir- teenth Indian Conference on Computer Vision, Graphics and Image Processing (ICVGIP 2022), Article 57, 9 ...

  35. [43]

    Wendland

    H. Wendland. Piecewise Polynomial, Positive Definite and Compactly Supported Radial Functions of Minimal Degree.Advances in Computational Mathematics, 4:389–396, 1995. https://doi.org/10.1007/BF02123482

  36. [44]

    J. P. Lewis. Algorithms for Solid Noise Synthesis.ACM SIGGRAPH Computer Graphics, 23(3):263–270, 1989.https://doi.org/10.1145/74334.74360 39

  37. [45]

    Lagae, S

    A. Lagae, S. Lefebvre, G. Drettakis, and P. Dutre. Procedural Noise Using Sparse Ga- bor Convolution.ACM Transactions on Graphics, 28(3), 2009.https://graphics.cs. kuleuven.be/publications/LLDD09PNSGC/

  38. [46]

    Kormann, C

    K. Kormann, C. Lasser, and A. Yurova. Stable Interpolation with Isotropic and Anisotropic Gaussians Using Hermite Generating Function.SIAM Journal on Scientific Computing, 41(6):A3839–A3859, 2019.https://doi.org/10.1137/19M1262449

  39. [47]

    Beatson, O

    R. Beatson, O. Davydov, and J. Levesley. Error Bounds for Anisotropic RBF Interpolation. Journal of Approximation Theory, 162(3):512–527, 2010.https://doi.org/10.1016/j. jat.2009.08.004

  40. [48]

    Casciola, D

    G. Casciola, D. Lazzaro, L. B. Montefusco, and S. Morigi. Shape Preserving Surface Reconstruction Using Locally Anisotropic RBF Interpolants.Computers & Mathematics with Applications, 51(8):1185–1198, 2006.https://doi.org/10.1016/j.camwa.2006.04. 002 40

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.