Pith. sign in

REVIEW 6 major objections 5 minor 1 cited by

Discrete Gaussian Vector Fields On Meshes

T0 review · 6 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper constructs discrete intrinsic Gaussian processes for vector fields on arbitrary two-dimensional meshes using the cotangent Laplacian and discrete exterior calculus, giving an exact discrete analogue of smooth manifold vector…

desk verdict A practical mesh-based vector GP construction that works in examples, but the 'exact discrete analogue' claim outruns the missing proof that the sharp operator preserves the inner product. read the letter →

arxiv 2507.20024 v1 pith:JCTAIZUI submitted 2025-07-26 stat.ME math.STstat.MLstat.TH

Michael Gillan (1) , Stefan Siegert (1) , Ben Youngman (1) ((1) University of Exeter) This is my paper
classification stat.MEmath.STstat.MLstat.TH MSC 62M3060G1558A1465N25
keywords GaussianprocessesvectorfieldsdiscreteexteriorcalculuscotangentLaplacianHelmholtz-HodgedecompositiontriangularmeshdownscalingMatérnkernels
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

This paper aims to show that intrinsic Gaussian processes for vector-valued fields, previously defined on smooth manifolds through the Hodge Laplacian, can be built directly on a triangular mesh without knowing the manifold's eigenfunctions. The authors construct discrete covariance kernels from the eigenvectors of the cotangent Laplacian, using discrete exterior calculus operators to build curl-free, divergence-free, and harmonic basis vector fields. If correct, this gives practitioners a single flexible prior for vector data such as winds and ocean currents on any two-dimensional surface, including the sphere, with boundary conditions and non-stationary length scales. The payoff is a practical statistical tool for downscaling gridded environmental data that respects the geometry and the physics of the underlying field.

What carries the argument

The load-bearing object is the cotangent Laplacian $L_c = \star_0^{-1} d_0^\top \star_1 d_0$, assembled from discrete exterior calculus operators on the mesh. Its generalized eigenpairs $(f_n, \lambda_n)$ solve $d_0^\top \star_1 d_0 f = \lambda \star_0 f$, and the eigenvectors replace the unknown Laplace-Beltrami eigenfunctions of the smooth manifold. Applying the exterior derivative $d_0$, the angle-weighted primal-primal sharp operator $(\cdot)^\#$, and the two-dimensional Hodge star (a positive $90^\circ$ rotation about the vertex normal) to these eigenvectors produces the curl-free and divergence-free vector basis. The functional calculus scaling $\Phi_{\nu,\kappa}(\lambda)$ from the scalar intrinsic Gaussian process literature then turns these bases into covariance matrices, making the construction mesh-agnostic, boundary-aware, and non-stationary.

What would settle it

Compute the discrete vector kernel on a sequence of refined meshes of a surface whose smooth Hodge-Laplacian spectrum is known, such as a flat torus or a sphere, and check whether the kernel's eigenvalues and eigenfields converge to the smooth intrinsic vector Gaussian process as the maximal edge length goes to zero; alternatively, measure the normal component of posterior samples on a Neumann no-flux boundary and see whether the small boundary flux vanishes under refinement.

Watch

Extended reading notes

Core claim

The central claim is that the smooth Hodge-compositional Matérn kernels of Robert-Nicoud et al. (2024) have an exact discrete analogue. On an arbitrary two-dimensional simplicial mesh embedded in Euclidean space, the paper defines vector-valued Gaussian process priors by taking eigenvectors $f_n$ of the cotangent Laplacian $L_c$, forming curl-free and divergence-free basis fields $f^d_n = (df_n/\sqrt{\lambda_n})^\#$ and $f^c_n = *(df_n/\sqrt{\lambda_n})^\#$, adding a harmonic basis from the zero eigenspace of the discrete Hodge-Laplacian, and assembling the kernel $K_v = \sigma_d^2 K_d + \sigma_c^2 K_c + \sigma_h^2 K_h$. The same construction accommodates no-flux boundary conditions via separate Dirichlet and Neumann eigenproblems and non-stationary length scales through a spatially varying $\kappa(s)$ in the spectral scaling; the paper demonstrates the resulting priors on global wind downscaling and ocean-current interpolation.

Load-bearing premise

The discrete sharp and Hodge-star operations used to turn edge-level derivatives into vertex-level vectors must faithfully reproduce the surface geometry and preserve inner products; if the angle-weighted interpolation distorts the metric or fails to converge as the mesh refines, the kernels are not the claimed intrinsic vector Gaussian processes.

Editorial extensions

If this is right

  • Any two-dimensional simplicial mesh, such as a sphere, torus, flat domain, or arbitrary surface, admits a discrete intrinsic vector Gaussian process prior without requiring closed-form manifold eigenfunctions.
  • No-flux boundary conditions are imposed by solving separate Dirichlet and Neumann eigenproblems for the curling and diverging components, and the examples confirm that posterior samples respect the boundary.
  • Non-stationary fields are handled by making the length-scale parameter $\kappa(s)$ a spatially varying function in the spectral scaling, and the paper infers latitude-dependent length scales consistent with the Rossby radius of deformation.
  • The decomposition into curl-free, divergence-free, and harmonic components lets the model isolate or down-weight physical components of the flow, as done when modeling high-altitude winds with only the divergence-free basis.
  • Downscaling on the globe and interpolation of ocean currents from sparse drifter observations both fall within the same formulation.

Reading between the lines

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

  • A natural extension not run in the paper is a mesh-refinement convergence study: as the maximal edge length tends to zero, the discrete vector kernels should approach the smooth Hodge-compositional kernels, and demonstrating this would strengthen the claim of an exact discrete analogue.
  • The boundary-flux artifact introduced by the sharp interpolation could likely be reduced by using a dual-primal sharp in the boundary layer, since the paper notes the two sharp definitions agree in the interior of a two-dimensional manifold.
  • Because the scalar kernel admits a sparse precision matrix while the vector kernel does not, a sparse-precision or inducing-point extension of the Hodge-compositional kernel would be the next practical step for scaling to very large meshes.
  • The same construction may transfer to edge-valued data with only a change of basis, connecting the mesh formulation to existing work on functions defined on mesh edges.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

6 major / 5 minor

Summary. The manuscript develops a discrete Gaussian process prior for vector fields on triangulated two-dimensional manifolds. It constructs covariance kernels from eigenpairs of the cotangent Laplacian, applies discrete exterior derivative and a primal-primal sharp operator to obtain curl-free basis fields, rotates these by a positive pi/2 rotation to obtain divergence-free basis fields, and adds a harmonic component. It then builds covariance matrices from truncated sums of outer products, with extensions to Dirichlet/Neumann boundary conditions and nonstationary length scales. Applications include downscaling ERA5 wind fields on a spherical mesh and ocean drifter currents in a bounded domain, with code provided.

Significance. If the paper's central 'exact discrete analogue' claim is established, this would be a useful and practical contribution: the construction is more flexible than existing graph-based vector GP methods, handles arbitrary meshes, boundary conditions, and nonstationarity, and the paper includes reproducible code and empirical demonstrations. The nonstationary example in Section 4.2 estimates the length-scale function from data rather than prescribing it, which is a genuine empirical contribution. At present, however, the geometric discretization step---the sharp operator and the rotation-based Hodge star---is not analyzed, so the intrinsic nature of the resulting kernels is asserted rather than demonstrated. The paper is of interest to the spatial statistics and machine learning communities, but the central theoretical claim needs substantial strengthening.

major comments (6)
  1. [Section 3.1] The paper establishes orthonormality of the edge-valued forms df_n/sqrt(lambda_n) with respect to the star_1 inner product, but the kernels Kd and Kc are sums of vertex-wise outer products of the sharped fields f^d_n and f^c_n. No result shows that Hirani's primal-primal sharp is an isometry from the edge-cochain space with the star_1 inner product to the vertex vector-field space with the appropriate L2 (mass-matrix) inner product. Consequently, the normalizing constant C = (1/A) sum Phi is not tied to the L2 norm of the discrete vector fields, and Kd and Kc are not shown to be the discrete analogues of the smooth Hodge-compositional kernels. This is the load-bearing step for the paper's central claim of an exact discrete analogue.
  2. [Section 3.1] The Helmholtz-Hodge decomposition interpretation is asserted but not verified. No discrete curl and divergence operators are defined on the vertex-valued vector fields produced by the sharp operator, so it is not established that f^d_n is curl-free or that f^c_n is divergence-free in any discrete sense. Without such operators and an explicit discrete analogue of the Hodge decomposition, the split into Kd, Kc, and Kh is not shown to be the discrete counterpart of the smooth decomposition used by Robert-Nicoud et al. (2024).
  3. [Section 3.2] The harmonic basis is not obtained from the same discrete exterior calculus construction except in special cases. The paper states that the DEC Hodge Laplacian has natural Neumann boundary conditions and that this route 'only works on unbounded manifolds where the 0-eigenspace exists, such as a torus'; for subsets of R2 it recommends defining a harmonic basis ad hoc, and for other bounded domains an outer boundary. This is not an exact discrete analogue and weakens the claim that the methodology applies to arbitrary 2D meshes with a prescribed harmonic component.
  4. [Section 3.3] The no-flux boundary condition is only approximate because the sharp interpolation introduces some small flux over the boundary, and the proposed fix is post-processing by removing the normal component at boundary vertices. This post-processing is not derived from the covariance kernel, and no statement is given about what kernel the post-processed field corresponds to or whether the Hodge decomposition remains valid. The empirical no-flux behavior in Section 4.3 therefore does not follow from the constructed GP prior.
  5. [Section 3.4] The nonstationary model is underspecified. Equation (3) defines Phi(lambda, s) pointwise, but the text only says that the scaling 'can be applied' to the curling and diverging basis fields; no formula is given for the resulting covariance matrix, and it is unclear whether the (i,j) entry uses Phi(lambda_n, s_i), Phi(lambda_n, s_j), a product, or a symmetrized average. Without this formula, the inference in Section 4.2 is not reproducible as written, and the claim that the method recovers the Rossby radius structure cannot be checked.
  6. [Sections 2.2-2.3 and 3.1] No convergence analysis or error estimates are provided for the vector kernels. The cited convergence of the cotangent Laplacian and of DEC operators does not by itself imply convergence of Kd and Kc to the smooth kernels, because the sharp interpolation and the rotation-based star are additional steps whose mesh-refinement behavior is not analyzed. The paper should either prove convergence under mesh refinement or explicitly soften the 'exact discrete analogue' claim to a heuristic discrete formulation.
minor comments (5)
  1. [Section 2.1, Eq. (2)] The placement of star_0 in Eq. (2), K = sigma^2/C F Phi(Λ) star_0 F^T, appears inconsistent with the orthogonality relation f_i^T star_0 f_j = delta_ij; as written the matrix is not obviously symmetric. If the intended scalar kernel is K = F Phi F^T, the formula should be corrected.
  2. [Section 3.1] The statement that applying the primal-primal sharp followed by a positive pi/2 rotation is equivalent to applying the Hodge star followed by a dual-primal sharp is not proved or cited; a short derivation or reference would clarify the construction.
  3. [Section 3.2] The caption of Figure 2 says the figure shows a 0-eigenspace for a subset of R2, while the text says this formulation only works on unbounded manifolds; the relationship between these two statements should be reconciled.
  4. [Section 4.2 and Figure 7] The text says there is a smaller length-scale near the equator than the tropics and that the length-scale decreases towards the poles; please clarify whether this is intended as two separate statements and connect the description to the plotted curve.
  5. [Section 3.3] The sentence 'This is similar to the eigenproblem posed in Bell and Hirani (2012, Section 11.1)' should give the exact relation or reproduce the relevant eigenproblem so that the Dirichlet treatment is fully specified.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the discrete vector kernels are formed by applying external DEC and smooth-manifold GP machinery, and no fitted parameter is relabeled as a prediction.

full rationale

The paper's discrete vector kernels are constructed by applying discrete exterior calculus operators to the eigenbasis of the cotangent Laplacian, following the smooth Hodge-compositional construction of Robert-Nicoud et al. (2024). The formulas for Kd, Kc, and Kh define the covariance structure directly; they do not fit a parameter to a subset of data and then present a forced quantity as a prediction. The inferred kappa(s) in Section 4.2 is a genuinely estimated parameter, and the Rossby-radius comparison is a qualitative consistency check rather than a fitted target. The citations to Borovitskiy et al., Robert-Nicoud et al., and Hirani are external sources, not self-citations by the present authors, and the central kernel definitions do not presuppose the paper's applications or results. The main vulnerability, that the angle-weighted primal-primal sharp is not proved to preserve the L2 inner product or Hodge decomposition, is a correctness or fidelity gap rather than circular reasoning: the paper asserts the analogy but does not reduce its conclusion to its own input by construction. No equation in the derivation chain is equivalent to its own output, and no fitted value is renamed as a prediction.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The construction rests on borrowed convergence results and a 2D-only rotation approximation; no new physical entities are introduced. The main uncharged premise is that the discrete sharp and rotation yield vertex-valued fields that preserve the smooth Hodge decomposition enough for GP inference.

free parameters (4)
  • Eigenbasis truncation L = 250 in practice
    Section 3.1: 'L is chosen to capture sufficient local detail, usually set to 250 in practice'. It controls the rank of the kernel and is not determined by data.
  • Smoothness nu = 1.5 in all examples
    Set to 1.5 in the wind and ocean applications; no sensitivity analysis is given.
  • Length-scale and variance hyperparameters = Ocean example: kappa_d=5.0, kappa_c=50.0, sigma_c=0.5, sigma_d=2.5; wind uses 'large kappa_c'
    Section 4.3 states these values without justification or cross-validation; Section 4.1 also fixes kappa_c and nu by hand.
  • Nonstationary kappa(s) low-rank weights = Posterior means from MCMC
    Section 4.2 infers weights for 20 RBF basis functions with standard normal priors; these are fitted to data and used to claim recovery of the Rossby radius structure.
assumptions (6)
  • domain assumption Smooth vector Matern kernels defined through eigenfunctions of the Hodge Laplacian are valid GP priors.
    Adopted from Robert-Nicoud et al. (2024); the whole discrete construction is presented as an analogue of this.
  • standard math Cotangent Laplacian eigenvalues and eigenvectors converge to Laplace-Beltrami as the mesh refines.
    Relied on in Sections 2.3 and 3.1 to justify using Lc in place of Delta_g.
  • standard math DEC operators d and star converge to smooth exterior calculus operators.
    Cited via Schulz and Tsogtgerel (2020); supports the claim that discrete kernels approximate smooth kernels.
  • standard math A 2D manifold with metric g is isometrically embeddable as a triangular mesh via Nash-Kuiper.
    Invoked in Section 2.2 to ensure the mesh embedding captures g.
  • ad hoc to paper Hodge star applied to vertex vectors can be replaced by a positive pi/2 rotation about the surface normal.
    Used in Section 3.1 to define f^c_n; the Discussion admits this restricts the method to 2D.
  • domain assumption Neumann boundary conditions are automatically generated by the discrete exterior derivative because there are no edges beyond the boundary.
    Used in Section 3.3 to construct no-flux boundaries; small boundary flux is admitted and post-processed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Discrete Gaussian Vector Fields On Meshes." pith.science (2026). https://pith.science/paper/JCTAIZUI

@misc{pith2026250720024,
  author       = {Pith},
  title        = {Pith review of: Discrete Gaussian Vector Fields On Meshes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JCTAIZUI}},
  note         = {Machine review of arXiv:2507.20024}
}
read the original abstract

Though the underlying fields associated with vector-valued environmental data are continuous, observations themselves are discrete. For example, climate models typically output grid-based representations of wind fields or ocean currents, and these are often downscaled to a discrete set of points. By treating the area of interest as a two-dimensional manifold that can be represented as a triangular mesh and embedded in Euclidean space, this work shows that discrete intrinsic Gaussian processes for vector-valued data can be developed from discrete differential operators defined with respect to a mesh. These Gaussian processes account for the geometry and curvature of the manifold whilst also providing a flexible and practical formulation that can be readily applied to any two-dimensional mesh. We show that these models can capture harmonic flows, incorporate boundary conditions, and model non-stationary data. Finally, we apply these models to downscaling stationary and non-stationary gridded wind data on the globe, and to inference of ocean currents from sparse observations in bounded domains.

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Mat\'ern Noise for Triangulation-Agnostic Flow Matching on Meshes

    cs.GR 2026-05 unverdicted novelty 7.0 of 10

    Proposes discretized Matérn process noise for triangulation-agnostic flow matching on meshes with PoissonNet denoiser, tested on elastic states and humanoid poses for meshes exceeding one million triangles.

Reference graph

Works this paper leans on

28 extracted references · 20 canonical work pages · cited by 1 Pith paper

  1. [1]

    , Hirani , A.N

    barticle Bell , N. , Hirani , A.N. : PyDEC : Software and Algorithms for Discretization of Exterior Calculus . ACM Trans. Math. Softw. 39 ( 1 ), 3 -- 1341 ( 2012 ) 10.1145/2382585.2382588 barticle

    arXiv 2012

  2. [2]

    , Trippe , B.L

    botherref Berlinghieri , R. , Trippe , B.L. , Burt , D.R. , Giordano , R. , Srinivasan , K. , \"O zg \"o kmen , T. , Xia , J. , Broderick , T. : Gaussian Processes at the Helm (Holtz): A More Fluid Model for Ocean Currents. arXiv (2023) botherref

    work page 2023

  3. [3]

    , Norgard , G

    barticle Bhatia , H. , Norgard , G. , Pascucci , V. , Bremer , P.-T. : The Helmholtz-Hodge Decomposition --- A Survey . IEEE Transactions on Visualization and Computer Graphics 19 ( 8 ), 1386 -- 1404 ( 2013 ) 10.1109/TVCG.2012.316 barticle

    doi:10.1109/tvcg.2012.316 2013

  4. [4]

    , Azangulov , I

    botherref Borovitskiy , V. , Azangulov , I. , Terenin , A. , Mostowsky , P. , Deisenroth , M.P. , Durrande , N. : Mat \'e rn Gaussian Processes on Graphs . arXiv (2021) botherref

    work page 2021

  5. [5]

    , Terenin , A

    botherref Borovitskiy , V. , Terenin , A. , Mostowsky , P. , Deisenroth , M.P. : Mat \'e rn Gaussian Processes on Riemannian Manifolds. arXiv (2023) botherref

    work page 2023

  6. [6]

    , De Goes , F

    bchapter Crane , K. , De Goes , F. , Desbrun , M. , Schr \"o der , P. : Digital geometry processing with discrete exterior calculus . In: ACM SIGGRAPH 2013 Courses , pp. 1 -- 126 . ACM , Anaheim California ( 2013 ). 10.1145/2504435.2504442 bchapter

    work page arXiv 2013

  7. [7]

    Copernicus Marine Service Information (CMEMS) : Global Ocean- in-Situ Near Real Time Observations of Ocean Currents

    botherref E.U. Copernicus Marine Service Information (CMEMS) : Global Ocean- in-Situ Near Real Time Observations of Ocean Currents. Marine Data Store (MDS) (2018). 10.48670/MOI-00041 botherref

    work page doi:10.48670/moi-00041 2018

  8. [8]

    , Raftery , A.E

    barticle Fuentes , M. , Raftery , A.E. : Model Evaluation and Spatial Interpolation by Bayesian Combination of Observations with Outputs from Numerical Models . Biometrics 61 ( 1 ), 36 -- 45 ( 2005 ) https://arxiv.org/abs/3695645 3695645 barticle

    work page arXiv 2005

Show all 28 references
  1. [9]

    : Atmosphere- Ocean Dynamics , 1 st edn

    bbook Gill , A. : Atmosphere- Ocean Dynamics , 1 st edn. International Geophysics Series , vol. 30 . Elsevier Science & Technology , Chantilly ( 1982 ) bbook

    1982

  2. [10]

    , Bell , B

    botherref Hersbach , H. , Bell , B. , Berrisford , P. , Biavati , G. , Hor \'a nyi , A. , Mu \ n oz Sabater , J. , Nicolas , J. , Peubey , C. , Radu , R. , Rozum , I. , Schepers , D. , Simmons , A. , Soci , C. , Dee , D. , Th \'e paut , J.-N. : ERA5 Hourly Data on Pressure Lev...

    2023 doi

  3. [11]

    , Bell , B

    botherref Hersbach , H. , Bell , B. , Berrisford , P. , Biavati , G. , Hor \'a nyi , A. , Mu \ n oz Sabater , J. , Nicolas , J. , Peubey , C. , Radu , R. , Rozum , I. , Schepers , D. , Simmons , A. , Soci , C. , Dee , D. , Th \'e paut , J.-N. : ERA5 Monthly Averaged Data on Pr...

    2023 doi

  4. [12]

    , Polthier , K

    barticle Hildebrandt , K. , Polthier , K. , Wardetzky , M. : On the convergence of metric and geometric properties of polyhedral surfaces . Geometriae Dedicata 123 ( 1 ), 89 -- 112 ( 2006 ) 10.1007/s10711-006-9109-5 barticle

    2006 doi

  5. [13]

    : Discrete Exterior Calculus

    botherref Hirani , A.N. : Discrete Exterior Calculus . PhD thesis (May 2003) botherref

    2003

  6. [14]

    , Terenin , A

    botherref Hutchinson , M. , Terenin , A. , Borovitskiy , V. , Takao , S. , Teh , Y.W. , Deisenroth , M.P. : Vector-Valued Gaussian Processes on Riemannian Manifolds via Gauge Independent Projected Kernels . arXiv (2021) botherref

    2021

  7. [15]

    : On C 1-isometric imbeddings

    barticle Kuiper , N.H. : On C 1-isometric imbeddings. I . Indagationes Mathematicae (Proceedings) 58 , 545 -- 556 ( 1955 ) 10.1016/S1385-7258(55)50075-8 barticle

    1955 doi

  8. [16]

    , Rue , H

    barticle Lindgren , F. , Rue , H. , Lindstr \"o m , J. : 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 (Statistical Methodology) 73 ( 4 ), 4...

    2011

  9. [17]

    , Bolin , D

    barticle Lindgren , F. , Bolin , D. , Rue , H. : The SPDE approach for Gaussian and non- Gaussian fields: 10 years and still running . Spatial Statistics 50 , 100599 ( 2022 ) 10.1016/j.spasta.2022.100599 https://arxiv.org/abs/2111.01084 arXiv:2111.01084 [stat] barticle

    2022

  10. [18]

    , Polthier , K

    barticle Pinkall , U. , Polthier , K. : Computing discrete minimal surfaces and their conjugates . Experimental Mathematics 2 ( 1 ), 15 -- 36 ( 1993 ) barticle

    1993

  11. [19]

    , Williams , C.K.I

    bbook Rasmussen , C.E. , Williams , C.K.I. : Gaussian Processes for Machine Learning . Adaptive Computation and Machine Learning . MIT Press , Cambridge, Mass ( 2006 ) bbook

    2006

  12. [20]

    , Krause , A

    botherref Robert-Nicoud , D. , Krause , A. , Borovitskiy , V. : Intrinsic Gaussian Vector Fields on Manifolds . arXiv (2024) botherref

    2024

  13. [21]

    : The Laplacian on a Riemannian Manifold : An Introduction to Analysis on Manifolds

    bbook Rosenberg , S. : The Laplacian on a Riemannian Manifold : An Introduction to Analysis on Manifolds . London Mathematical Society Student Texts . Cambridge University Press , Cambridge ( 1997 ). 10.1017/CBO9780511623783 bbook

    1997 doi

  14. [22]

    , Tsogtgerel , G

    barticle Schulz , E. , Tsogtgerel , G. : Convergence of Discrete Exterior Calculus Approximations for Poisson Problems . Discrete & Computational Geometry 63 ( 2 ), 346 -- 376 ( 2020 ) 10.1007/s00454-019-00159-x https://arxiv.org/abs/1611.03955 arXiv:1611.03955 [math] barticle

    2020 arXiv

  15. [23]

    : Discrete differential operators on polyhedral surfaces - convergence and approximation

    botherref Wardetzky , M. : Discrete differential operators on polyhedral surfaces - convergence and approximation. PhD thesis (2007) botherref

    2007

  16. [24]

    , Mathur , S

    bchapter Wardetzky , M. , Mathur , S. , K \"a lberer , F. , Grinspun , E. : Discrete Laplace operators: No free lunch . In: ACM SIGGRAPH ASIA 2008 Courses on - SIGGRAPH Asia '08 , pp. 1 -- 5 . ACM Press , Singapore ( 2008 ). 10.1145/1508044.1508063 bchapter

    2008

  17. [25]

    : Stochastic Processes in Several Dimensions

    barticle Whittle , P. : Stochastic Processes in Several Dimensions . Bulletin of the International Statistical Institute 40 ( 2 ), 974 -- 994 ( 1963 ) barticle

    1963

  18. [26]

    , Krause , A

    botherref Wyrwal , K. , Krause , A. , Borovitskiy , V. : Residual Deep Gaussian Processes on Manifolds . arXiv (2024) botherref

    2024

  19. [27]

    , Youngman , B.D

    barticle Xiong , X. , Youngman , B.D. , Economou , T. : Data fusion with Gaussian processes for estimation of environmental hazard events . Environmetrics 32 ( 3 ), 2660 ( 2021 ) 10.1002/env.2660 barticle

    2021 doi

  20. [28]

    , Borovitskiy , V

    botherref Yang , M. , Borovitskiy , V. , Isufi , E. : Hodge- Compositional Edge Gaussian Processes . arXiv (2024) botherref

    2024

Pith tools

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