Pith. sign in

REVIEW 4 major objections 6 minor 26 references

Generalized Finite Difference Method for Solving Stochastic Diffusion Equations

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims that the generalized finite difference method—a meshless technique based on Taylor expansion and moving least squares—solves stochastic diffusion equations with mean-square consistency, conditional stability, and…

desk verdict A useful first application of GFDM to stochastic diffusion equations, but the stability and convergence proofs have a real gap around the sign of the stencil coefficients. read the letter →

arxiv 2411.14333 v1 pith:FXGBQCQW submitted 2024-11-21 math.NA cs.NA

Faezeh Nassajian Mojarrad This is my paper · ORCID
classification math.NAcs.NA MSC 65M0665M1260H1535R60
keywords stochasticdiffusionequationgeneralizedfinitedifferencemethodmeshlessschememean-squareconvergencestabilitymovingleastsquarespartialdifferential
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 introduces the generalized finite difference method for stochastic diffusion equations. It claims that the fully explicit scheme, in which the Laplacian at each scattered node is approximated by a weighted combination of neighboring values, is consistent, conditionally stable, and convergent in mean square, provided the diffusion coefficient, time step, and stencil coefficients satisfy $0 \leq \rho\Delta t \max_c \theta_c \leq 1$. If the claim holds, stochastic diffusion problems on irregular domains can be solved without generating a mesh, using only a cloud of points. The paper validates the method with one-, two-, and three-dimensional examples.

What carries the argument

The machinery is the weighted moving-least-squares Laplace stencil: around each central node $x_c$, second-order Taylor expansions to all star neighbors are multiplied by a positive symmetric weight, squared, and minimized; the resulting normal equations produce the approximation $-\theta_c u_c + \sum_i \theta_i u_i = 0$ for $\nabla^2 v$. The $\theta_i$ are recovered through a Cholesky decomposition of the moment matrix $H^{(d)}$, so they encode the local point geometry. Stability then hinges on the nonnegative combination $\theta_c = \sum_i \theta_i$ and the step-size bound $\rho\Delta t \max_c \theta_c \leq 1$, which makes the amplification factor $|1 - \rho\Delta t \theta_c| + \rho\Delta t \theta_c = 1$.

What would settle it

On an irregular cloud where a central node has two close neighbors on one side and one farther neighbor on the other, compute the Laplace stencil weights $\theta_i$ from the moving-least-squares normal equations; if any $\theta_i < 0$, then $\sum |\theta_i| > \theta_c$ and the energy bound of Theorem 2 cannot hold. Then check whether the claimed stability threshold is still respected by observing divergence when $\rho\Delta t \max_c \theta_c$ exceeds 1 on that cloud.

Watch

Extended reading notes

Core claim

The central claim, stated as Theorem 3, is that scheme (9) converges to the solution of (1) in the sup-norm in mean square whenever $0 \leq \rho\Delta t \max_c \theta_c \leq 1$. The convergence proof combines consistency (Theorem 1), which follows from the Taylor-series truncation error of the moving-least-squares Laplace approximation, with conditional stability (Theorem 2), and applies the stochastic Lax–Richtmyer theorem. The scheme advances each node value by the Euler–Maruyama update $u_c^{k+1} = u_c^k + \rho\Delta t(-\theta_c u_c^k + \sum_i \theta_i u_i^k) + \mu u_c^k (W((k+1)\Delta t) - W(k\Delta t))$, where the $\theta_i$ are the geometry-dependent stencil weights.

Load-bearing premise

The stability proof only goes through if the stencil weights that the method produces are all nonnegative, and if the time step is small enough relative to the diffusion coefficient and these weights; the paper does not establish the former for arbitrary clouds, nor does it report the parameter values in its own experiments.

Editorial extensions

If this is right

  • If correct, the scheme provides a meshless way to solve stochastic diffusion equations on domains where mesh generation is impractical, such as evolving or highly irregular geometries.
  • The step-size condition $0 \leq \rho\Delta t \max_c \theta_c \leq 1$ gives a computable, CFL-like stability threshold that can guide time-step selection once the stencil weights are known.
  • The same Taylor-expansion and moving-least-squares machinery applies in one, two, and three dimensions, so the method is dimension-agnostic in its construction.
  • Because convergence is measured in mean square in the sup-norm, the method is directly usable for estimating solution expectations, which is the quantity most SPDE applications need.

Reading between the lines

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

  • The nonnegativity of $\theta_i$ is likely to fail on strongly anisotropic clouds; if so, a bounded-falloff weight function or a larger star could restore it, which would be a testable extension of the stability proof.
  • One could convert the stability condition into an explicit step-size selector by estimating $\max_c \theta_c$ in terms of local spacing and star size, turning the theoretical threshold into a practical refinement rule.
  • The same stencil-based approximation could be applied to other stochastic PDEs, such as advection or conservation laws, with the divergence operator built from the same weighted Taylor expansions.
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

4 major / 6 minor

Summary. The paper proposes a generalized finite difference method (GFDM) for the stochastic diffusion equation (1), combining moving-least-squares spatial stencels with an explicit Euler-Maruyama-type time discretization. It claims pointwise consistency in mean square (Theorem 1), conditional stability under the condition 0 ≤ ρΔt max_c θ_c ≤ 1 (Theorem 2), and convergence via a stochastic Lax-Richtmyer theorem (Theorem 3). Numerical experiments in one, two, and three spatial dimensions compare Monte Carlo means of the numerical solution with exact expected solutions. The central claim is that the method is a consistent, conditionally stable, and convergent meshless solver for this SPDE; however, the stability proof contains a sign-related gap, the convergence theorem is asserted rather than proved, and the numerical section does not report the parameters needed to verify the hypotheses.

Significance. If the stability and convergence results were established rigorously, the paper would make a useful contribution: it extends GFDM, a well-known meshless method for deterministic PDEs, to stochastic diffusion equations, and it provides explicit stencil formulations in 1D, 2D, and 3D. The consistency proof uses Itô isometry and a Taylor-expansion argument, and the numerical section covers three spatial dimensions. The main strength is the novelty of combining GFDM with Monte Carlo simulation for SPDEs. However, the central stability and convergence claims are not proven as stated, because the proof silently requires nonnegative stencil coefficients and a quantitative consistency argument that are absent. The numerical validation tests only the error of the mean, not the mean-square convergence asserted in the theorems, and omits the parameter values needed to check the stability condition. These issues are load-bearing, so the paper in its current form does not justify its main theorems.

major comments (4)
  1. [Section 3, Theorem 2 and the bound preceding Eq. (15)] The proof of Theorem 2 replaces Σ_i |θ_i| by θ_c when bounding E|u^{k+1}_c|². This step is valid only if every stencil coefficient θ_i is nonnegative, and the stated condition 0 ≤ ρΔt max_c θ_c ≤ 1 also requires θ_c ≥ 0 for every central node. The paper never proves that the moving-least-squares construction of Section 2 yields nonnegative θ_i on irregular clouds, and in general such stencils can have mixed signs. If some θ_i are negative, then Σ_i |θ_i| can be substantially larger than θ_c, so the amplification factor can exceed one even when ρΔt max_c θ_c ≤ 1, invalidating the stability conclusion. The theorem should either include an explicit nonnegativity hypothesis and a proof or reference for it, or the stability proof must be replaced by one that does not require this assumption.
  2. [Section 3, Theorem 3] Convergence is asserted solely by citing a 'stochastic Lax-Richtmyer theorem' from Roth (1989). The paper does not verify the hypotheses of that theorem (for example, well-posedness of the SPDE, the precise norm in which consistency and stability are required, and the equivalence of consistency plus stability with convergence in mean square for this class of equations). Theorem 3 therefore does not follow from Theorems 1 and 2 as written. The authors should either state a precise version of the stochastic Lax-Richtmyer theorem and verify all of its conditions, or prove convergence directly by substituting the local consistency and stability bounds into the error equation.
  3. [Section 4, L2-error definition and Tables 1–4] The reported L2-error is computed as √(1/(N_t N)) Σ_{k,c} |v(x_c,t_k) − E(u^k_c)|², which measures only the error of the Monte Carlo mean. Definition 3 and Theorem 3 concern E∥u^{k+1} − v^{k+1}∥², i.e., strong mean-square convergence. The experiments therefore do not validate the claimed convergence. In addition, the paper does not report Δt, θ_c, or the star configurations for the irregular clouds, so it is impossible to check whether the stability condition 0 ≤ ρΔt max_c θ_c ≤ 1, or the nonnegativity assumption required by Theorem 2, is satisfied in the presented runs.
  4. [Section 3, Theorem 1 and Appendix A] The consistency proof expresses the stencil error as a sum of higher-order Taylor terms multiplied by θ_i and concludes that the expression tends to zero as δ → 0. This conclusion requires a uniform bound or an explicit scaling for θ_i in terms of δ; since the θ_i have dimension 1/length², they may grow without bound as the star shrinks. The paper should state the standard GFDM regularity assumptions (for example, from Benito et al. 2007) that make this estimate quantitative, rather than relying on an informal 'tends to zero' assertion.
minor comments (6)
  1. [Section 2, after Eq. (1)] The text says 'µ is the additive noise', but the term µ v dW(t) is multiplicative noise; if additive noise is intended, the equation should read µ dW(t).
  2. [Section 3, Theorem 1] The theorem assumes only 'continuously differentiable' ψ, but the proof uses Taylor series with third- and higher-order derivatives; the hypothesis should require ψ ∈ C^4 or an equivalent regularity condition.
  3. [Section 4, Table 1 and Tables 2–3] No convergence rates are reported; the 1D table shows only three refinement levels, and the 2D tables compare different clouds rather than a systematic refinement sequence, so the statement that 'error decay fulfils the expected behavior' is not quantitatively supported.
  4. [Figure 4 and Figure 8] The captions say the stability condition is not satisfied, but the values of Δt and θ_c are not reported, making the instability demonstration unverifiable.
  5. [References] Some reference entries are incomplete or contain typographical errors; in particular, the entries for Mojarrad and Soheili (2024) and Veiga et al. (2024) lack full publication data, and the Veiga entry lists the same author twice.
  6. [Throughout] There are several typographical and formatting issues, such as 'T able 1', 'the rest of nodes are', and 'Summing' capitalized mid-sentence; these should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivation rests on external GFDM consistency results and the classical Lax-Richtmyer theorem; the few self-citations are background only and carry no load-bearing argument.

full rationale

The paper's derivation chain is self-contained in the sense required by the circularity test. Theorem 1 (consistency) is proved by direct Taylor expansion of the local truncation error, with the spatial GFDM approximation imported from Benito et al. (2001, 2007), which are independent prior works, not self-citations. Theorem 2 (stability) estimates E|u^{k+1}_c|^2 directly from the scheme and uses the stated condition 0 <= rho*dt*max_c(theta_c) <= 1; Theorem 3 then invokes the standard stochastic Lax-Richtmyer theorem from Roth (1989), an external result. The author's self-citations (Mojarrad 2024; Mojarrad and Soheili 2024; Veiga et al. 2024) appear only in the introduction as examples of SDE/SPDE methods or in the conclusion as future work; none of them supplies the premise of Theorems 1-3. No parameter is fitted to a data subset and then renamed as a prediction: the numerical experiments compare the scheme against analytically known expected solutions, and the only tunable ingredient, the weight function, is chosen a priori as w_i = 1/delta_i^3. There is a genuine mathematical gap in the proof of Theorem 2, because the inequality bounding the cross terms by (|1-rho*dt*theta_c| + |rho*dt*theta_c|)^2 uses sum(|theta_i|) = theta_c and therefore silently assumes all theta_i are nonnegative; on irregular clouds the moving-least-squares stencils can have mixed signs. However, that is a correctness/stability-assumption issue, not circularity: the missing assumption is not the target result and is not established by citing the present paper or its authors. The central claim therefore has independent mathematical content.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The paper's central claims rest on standard stochastic calculus and a set of unverified conditions: smoothness of the solution, invertibility of the GFDM star matrices, nonnegativity of the stencil coefficients, and the applicability of an external Lax-Richtmyer theorem. The free parameters (weight exponent, star size) are chosen by hand and not analyzed.

free parameters (2)
  • Weight function exponent n = 3
    The potential weight w_i = 1/delta_i^n is used with n=3 in all numerical tests (Section 4); no sensitivity analysis is given, and the choice affects the stencil coefficients theta_i.
  • Star size M = not reported
    The number of nearest neighbors per star is not specified; Section 2 only says the closest nodes are chosen (Jensen 1972), leaving an implementation-dependent parameter that changes H and theta_i.
assumptions (5)
  • standard math Itô isometry for the Wiener integral
    Used in Theorem 1 to convert the expected square of the stochastic increment into a time integral.
  • domain assumption Exact solution is sufficiently smooth for Taylor expansion
    Section 2 introduces the Taylor expansion for 'sufficiently differentiable function v(x)', which is assumed throughout the consistency proof.
  • domain assumption Star matrices H^(2), H^(5), H^(9) are positive definite and invertible
    Section 2 uses Cholesky decomposition of H, which requires positive definiteness; this is not proven for all irregular node configurations.
  • domain assumption Stochastic Lax-Richtmyer theorem from Roth (1989) applies
    Theorem 3 invokes it without stating hypotheses or verifying them for scheme (9).
  • ad hoc to paper Stencil coefficients theta_i are nonnegative
    Theorem 2's proof requires sum(|theta_i|) <= theta_c; the paper only states 0 <= rho*dt*max_c(theta_c) <= 1, which does not ensure theta_i >= 0.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Generalized Finite Difference Method for Solving Stochastic Diffusion Equations." pith.science (2026). https://pith.science/paper/FXGBQCQW

@misc{pith2026241114333,
  author       = {Pith},
  title        = {Pith review of: Generalized Finite Difference Method for Solving Stochastic Diffusion Equations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FXGBQCQW}},
  note         = {Machine review of arXiv:2411.14333}
}
read the original abstract

Stochastic diffusion equations are crucial for modeling a range of physical phenomena influenced by uncertainties. We introduce the generalized finite difference method for solving these equations. Then, we examine its consistency, stability and convergence in mean-square, showing that the proposed method preserves stability and demonstrates favorable convergence characteristics under suitable assumptions. In order to validate the methodology, we present numerical results in one-, two-, and three-dimensional space domains.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 16 canonical work pages

  1. [1]

    IMA Journal of Numerical Analysis 40(1), 247–284 (2020) https://doi.org/10.1093/imanum/dry060

    Anton, R., Cohen, D., Quer-Sardanyons, L.: A fully discrete approximation of the one- dimensional stochastic heat equation. IMA Journal of Numerical Analysis 40(1), 247–284 (2020) https://doi.org/10.1093/imanum/dry060

    work page doi:10.1093/imanum/dry060 2020

  2. [2]

    Applied Mathematical Modelling 25(12), 1039–1053 (2001) https://doi.org/10.1016/S0307-904X(01)00029-4

    Benito, J.J., Urena, F., Gavete, L.: Influence of several factors in the generalized finite difference method. Applied Mathematical Modelling 25(12), 1039–1053 (2001) https://doi.org/10.1016/S0307-904X(01)00029-4

    work page doi:10.1016/s0307-904x(01)00029-4 2001

  3. [3]

    Journal of computational and applied mathematics 209(2), 208–233 (2007) https://doi.org/10.1016/j.cam.2006.10.090 19

    Benito, J.J., Urena, F., Gavete, L.: Solving parabolic and hyperbolic equations by the generalized finite difference method. Journal of computational and applied mathematics 209(2), 208–233 (2007) https://doi.org/10.1016/j.cam.2006.10.090 19

    doi:10.1016/j.cam.2006.10.090 2007

  4. [4]

    Journal of Computational and Applied Mathematics 205(2), 912–922 (2007) https://doi.org/10.1016/j.cam.2006.03.038

    Buckwar, E., Winkler, R.: Improved linear multi-step methods for stochastic ordinary differential equations. Journal of Computational and Applied Mathematics 205(2), 912–922 (2007) https://doi.org/10.1016/j.cam.2006.03.038

    work page doi:10.1016/j.cam.2006.03.038 2007

  5. [5]

    Mathematische annalen 100(1), 32–74 (1928) https://doi

    Courant, R., Friedrichs, K., Lewy, H.: ¨Uber die partiellen differenzengleichungen der mathematischen physik. Mathematische annalen 100(1), 32–74 (1928) https://doi. org/10.1007/BF01448839

    doi:10.1007/bf01448839 1928

  6. [6]

    Applied Numerical Mathematics 157, 135–158 (2020) https://doi.org/10.1016/j.apnum

    Chen, Z., Gan, S., Wang, X.: A full-discrete exponential euler approximation of the invariant measure for parabolic stochastic partial differential equations. Applied Numerical Mathematics 157, 135–158 (2020) https://doi.org/10.1016/j.apnum. 2020.05.008

    doi:10.1016/j.apnum 2020

  7. [7]

    Journal of Differential Equations263(7), 3687–3713 (2017) https://doi.org/10.1016/j.jde.2017.05.002

    Cui, J., Hong, J., Liu, Z.: Strong convergence rate of finite difference approximations for stochastic cubic schr¨ odinger equations. Journal of Differential Equations263(7), 3687–3713 (2017) https://doi.org/10.1016/j.jde.2017.05.002

    doi:10.1016/j.jde.2017.05.002 2017

  8. [8]

    Transfer, Part B: Fundamentals 65(2), 129–154 (2014) https://doi.org/10.1080/ 10407790.2013.849979

    arXiv 2014

Show all 26 references
  1. [9]

    The European Physical Journal Plus 134, 1–20 (2019) https://doi.org/10.1140/epjp/i2019-12786-7

    Fu, Z.J., Tang, Z.C., Zhao, H.T., Li, P.W., Rabczuk, T.: Numerical solutions of the coupled unsteady nonlinear convection-diffusion equations based on generalized finite difference method. The European Physical Journal Plus 134, 1–20 (2019) https://doi.org/10.1140/epjp/i2019-12786-7

    2019 doi

  2. [10]

    Journal of Computational and Applied Mathematics 318, 378–387 (2017) https://doi.org/10.1016/j.cam.2016.07.025

    Gavete, L., Ure˜ na, F., Benito, J.J., Garc ´ ıa, A., Ure˜ na, M., Salete, E.: Solving second order non-linear elliptic partial differential equations using generalized finite differ- ence method. Journal of Computational and Applied Mathematics 318, 378–387 (2017) https://doi...

    2017 doi

  3. [11]

    World Applied Sciences Journal 21, 95–100 (2013) https://doi.org/10.5829/idosi.wasj.2013.21.am.1396

    Izadian, J., Ranjbar, N., Jalili, M.: The generalized finite difference method for solving elliptic equation on irregular mesh. World Applied Sciences Journal 21, 95–100 (2013) https://doi.org/10.5829/idosi.wasj.2013.21.am.1396

    2013 doi

  4. [12]

    Computers & Structures 2(1-2), 17–29 (1972) https://doi.org/10.1016/0045-7949(72)90020-X

    Jensen, P.S.: Finite difference techniques for variable grids. Computers & Structures 2(1-2), 17–29 (1972) https://doi.org/10.1016/0045-7949(72)90020-X

    1972 doi

  5. [13]

    CBMS-NSF Regional Conference Series in Applied Mathematics, Series Number 83, SIAM, Philadelphia (2011)

    Jentzen, A., Kloeden, P.E.: Taylor Approximations for Stochastic Partial Differential Equations. CBMS-NSF Regional Conference Series in Applied Mathematics, Series Number 83, SIAM, Philadelphia (2011). https://doi.org/10.1137/1.9781611972016

    2011 doi

  6. [14]

    IMA Journal of Numerical Analysis 40(2), 1005–1050 (2020) https://doi.org/10.1093/imanum/drz009 20

    Jentzen, A., Puˇ snik, P.: Strong convergence rates for an explicit numerical approx- imation method for stochastic evolution equations with non-globally lipschitz continuous nonlinearities. IMA Journal of Numerical Analysis 40(2), 1005–1050 (2020) https://doi.org/10.1093/iman...

    2020 doi

  7. [15]

    https://doi.org/10.1007/978-3-662-12616-5

    Springer, Heidelberg (1992). https://doi.org/10.1007/978-3-662-12616-5

    1992 doi

  8. [16]

    Engineering Analysis with Boundary Elements 80, 58–71 (2017) https://doi.org/10.1016/j.enganabound.2017.03.012

    Li, P.W., Fan, C.M.: Generalized finite difference method for two-dimensional shallow water equations. Engineering Analysis with Boundary Elements 80, 58–71 (2017) https://doi.org/10.1016/j.enganabound.2017.03.012

    2017 doi

  9. [17]

    Applied Mathematics Letters 114(106896) (2021) https://doi

    Li, P.W.: Space–time generalized finite difference nonlinear model for solving unsteady burgers’ equations. Applied Mathematics Letters 114(106896) (2021) https://doi. org/10.1016/j.aml.2020.106896

    2021

  10. [18]

    IMA Journal of Numerical Analysis 43(2), 1001–1035 (2023) https://doi.org/10.1093/imanum/drab107

    Li, Y., Mao, X., Song, Q., Wu, F., Yin, G.: Strong convergence of euler–maruyama schemes for mckean–vlasov stochastic differential equations under local lipschitz conditions of state variables. IMA Journal of Numerical Analysis 43(2), 1001–1035 (2023) https://doi.org/10.1093/i...

    2023 doi

  11. [19]

    Computers & Structures 11(1-2), 83–95 (1980) https://doi.org/10.1016/0045-7949(80)90149-2

    Liszka, T., Orkisz, J.: The finite difference method at arbitrary irregular grids and its application in applied mechanics. Computers & Structures 11(1-2), 83–95 (1980) https://doi.org/10.1016/0045-7949(80)90149-2

    1980 doi

  12. [20]

    arXiv preprint, arXiv:2407.19350 (2024) https://doi.org/10.48550/arXiv

    Mojarrad, F.N.: A new approximation method for solving stochastic differential equations. arXiv preprint, arXiv:2407.19350 (2024) https://doi.org/10.48550/arXiv. 2407.19350

    2024 doi arXiv

  13. [21]

    Mojarrad, F.N., Soheili, A.R.: Approximation of stochastic advection-diffusion equations with predictor–corrector methods (2024)

    2024

  14. [22]

    Halle-Wittenberg, Halle (1989)

    1989

  15. [23]

    Computational and Applied Mathematics 40(4) (2021) https: //doi.org/10.1007/s40314-021-01495-6

    Shahmoradi, M., Ahmadian, D., Ranjbar, M.: Mean-square stability of 1.5 strong convergence orders of diagonally drift runge–kutta methods for a class of stochastic differential equations. Computational and Applied Mathematics 40(4) (2021) https: //doi.org/10.1007/s40314-021-01495-6

    2021 doi

  16. [24]

    arXiv preprint, arXiv:2408.05081 (2024) https://doi.org/10

    Veiga, M.H., Mojarrad, F.N., Mojarrad, F.N.: Learning a robust shape parameter for rbf approximation. arXiv preprint, arXiv:2408.05081 (2024) https://doi.org/10. 48550/arXiv.2408.05081

    2024 doi arXiv

  17. [25]

    Yang, X., Zhao, W.: Strong convergence of a fully discrete scheme for multiplica- tive noise driving spdes with non-globally lipschitz continuous coefficients. Numer. Math. Theory Methods Appl 14, 1085–1109 (2021) https://doi.org/10.4208/nmtma. OA-2020-0143

    2021 doi

  18. [26]

    Zhao, W., Gunzburger, M.: Auxiliary equations approach for the stochastic unsteady navier–stokes equations with additive random noise. Numer. Math. Theory Methods 21 Appl 13, 1–26 (2020) https://doi.org/10.4208/nmtma.OA-2019-0055 22

    2020 doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.