pith. sign in

arxiv: 2605.21046 · v1 · pith:M6UTJ6PZnew · submitted 2026-05-20 · 🧮 math.NA · cs.NA

Stochastic Galerkin and Monte-Carlo methods for parabolic problems: Numerical performance of variational matrix-free approximations

Moataz Dawor , Nils Margenberg , Markus Bause This is my paper

Pith reviewed 2026-05-21 02:02 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords stochastic GalerkinMonte Carloparabolic problemsfinite elementsmatrix-freegeometric multigridpreconditioning
0
0 comments X

The pith

Stochastic Galerkin discretization outperforms Monte-Carlo sampling for parabolic problems with random variables in direct numerical comparisons.

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

The paper develops a stochastic Galerkin discretization for parabolic problems with random variables that embeds a slabwise finite element approximation in space and time. Large linear systems from this discretization are solved with GMRES iterations block-preconditioned by geometric multigrid using a local Vanka smoother, all within a matrix-free implementation. The identical solver technology is applied to a Monte-Carlo sampling strategy built on the same space-time finite element formulation. Numerical experiments on selected test problems track discretization convergence and algebraic solver statistics, showing the Galerkin approach requires fewer resources overall.

Core claim

The authors establish through numerical evaluation that the stochastic Galerkin method with variational matrix-free approximations achieves superior performance over Monte-Carlo sampling when both techniques employ consistent space-time finite element discretizations and the same block-preconditioned GMRES-GMG solver on random parabolic problems.

What carries the argument

Stochastic Galerkin discretization with slabwise space-time finite elements, solved by GMRES block-preconditioned by geometric multigrid with local Vanka smoother in a unified matrix-free framework.

If this is right

  • The matrix-free variational framework makes high-dimensional systems from random variables computationally tractable for both methods.
  • Discretization convergence rates remain comparable while algebraic solver costs diverge in favor of the Galerkin formulation.
  • Statistics of GMRES iterations and multigrid performance provide quantitative evidence for the efficiency gain.
  • The unified implementation in a single software architecture enables direct side-by-side evaluation without implementation bias.

Where Pith is reading between the lines

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

  • The observed advantage may motivate similar variational embeddings for other stochastic time-dependent equations such as hyperbolic or parabolic systems with nonlinear random coefficients.
  • Matrix-free geometric multigrid with Vanka smoothing could be adapted to related uncertainty-quantification tasks outside parabolic models.
  • Extending the comparison to time-adaptive or goal-oriented error estimators would test whether the performance gap persists under more sophisticated discretizations.

Load-bearing premise

The specific parabolic test problems, random variable distributions, and chosen discretization parameters used in the numerical experiments are representative of the broader class of random parabolic problems.

What would settle it

Repeating the full performance comparison on a parabolic problem with a qualitatively different random field, such as one with non-Gaussian statistics or stronger spatial correlation, and finding that Monte-Carlo statistics improve relative to Galerkin would falsify the observed superiority.

Figures

Figures reproduced from arXiv: 2605.21046 by Markus Bause, Moataz Dawor, Nils Margenberg.

Figure 1
Figure 1. Figure 1: Stochastic-refinement error diagnostics. [PITH_FULL_IMAGE:figures/full_fig_p013_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Stochastic-refinement rate diagnostics. The full-error rates increase over this range, which indicates that the stochastic truncation error is reduced more efficiently as the polynomial space approaches the exact degree. The variance rates are larger than the full-error rates, showing that the fluctuation component is resolved particularly efficiently by stochastic enrichment. The mean-rate plot again refl… view at source ↗
Figure 3
Figure 3. Figure 3: Reported work proxy W(pξ) for the finite polynomial stochastic-refinement benchmark [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Average FGMRES iterations per time slab for the stochastic-refinement study as a function of the chaos [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Timing breakdown for the stochastic Galerkin solves in the stochastic-refinement study. [PITH_FULL_IMAGE:figures/full_fig_p016_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Monte-Carlo mean-error diagnostics as a function of the sample number. The curves illustrate the non [PITH_FULL_IMAGE:figures/full_fig_p023_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Monte-Carlo variance-error diagnostics as a function of the sample number. The variance estimator is more [PITH_FULL_IMAGE:figures/full_fig_p024_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Combined work-accuracy comparison for mean and variance errors. The upper panel shows the total mean [PITH_FULL_IMAGE:figures/full_fig_p026_8.png] view at source ↗
read the original abstract

Stochastic Galerkin methods offer unexplored potential for the numerical simulation of parabolic problems with random variables, in particular if they are combined with variational discretizations of the space and time variables. Due to the high dimensionality, the solution of the arising algebraic systems do not become feasible without efficient solvers, preconditioners, and software architectures. A stochastic Galerkin discretization with an embedded slabwise finite element approximation of the space and time variables is proposed and analyzed numerically. For solving the linear systems, GMRES iterations are block-preconditioned by a geometric multigrid (GMG) technique using a local Vanka smoother for the space-time subsystems. Monte-Carlo methods are also used for solving random parabolic problems and studied here for the purpose of comparison. The Monte-Carlo approach is built on the space-time finite element formulation together with the GMRES-GMG solver technology. All algorithms have been implemented in a unified matrix-free framework based on the deal.II software library. Comparative numerical evaluations illustrate the performance properties of both approaches, including convergence of the discretizations and statistics of the algebraic solver. Superiority of the stochastic Galerkin approach is observed.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The manuscript proposes a stochastic Galerkin discretization for parabolic PDEs with random variables, combined with a slabwise finite-element approximation in space and time. The resulting algebraic systems are solved by GMRES preconditioned by geometric multigrid using a local Vanka smoother, all realized in a matrix-free framework inside deal.II. An analogous Monte-Carlo method is constructed on the identical space-time finite-element formulation and solver technology. Numerical experiments report discretization convergence and algebraic-solver statistics, from which the authors conclude that the stochastic Galerkin approach is superior.

Significance. Should the observed performance advantage prove robust, the combination of variational space-time discretizations with block-preconditioned geometric multigrid in a matrix-free setting would constitute a practical advance for uncertainty quantification in time-dependent problems. The unified implementation and explicit solver statistics are concrete strengths that aid reproducibility.

major comments (1)
  1. [Numerical experiments] Numerical experiments section: the reported superiority of the stochastic Galerkin method rests on a single fixed collection of test problems, random-variable distributions, stochastic dimension, slabwise mesh, and coefficient contrast. No systematic variation of these quantities (e.g., increasing stochastic dimension, shortening correlation length, or raising contrast) is presented, so it is unclear whether the performance gap is intrinsic to the variational formulation or an artifact of the chosen block structure and parameter regime.
minor comments (1)
  1. [Abstract] The abstract states that 'superiority of the stochastic Galerkin approach is observed' without indicating the precise error norms or iteration-count metrics that support the claim.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful reading of the manuscript and for highlighting its potential significance. We address the major comment below and describe the revisions we intend to incorporate.

read point-by-point responses

  1. Referee: Numerical experiments section: the reported superiority of the stochastic Galerkin method rests on a single fixed collection of test problems, random-variable distributions, stochastic dimension, slabwise mesh, and coefficient contrast. No systematic variation of these quantities (e.g., increasing stochastic dimension, shortening correlation length, or raising contrast) is presented, so it is unclear whether the performance gap is intrinsic to the variational formulation or an artifact of the chosen block structure and parameter regime.

    Authors: We agree that the current numerical study is confined to a representative but fixed parameter regime and that additional variation would help establish the robustness of the observed performance difference. The experiments were chosen to enable a direct, apples-to-apples comparison of the stochastic Galerkin and Monte-Carlo approaches within the same space-time finite-element and matrix-free GMRES-GMG framework. In the revised manuscript we will add a new subsection containing results for increased stochastic dimensions (up to d=8) and for shorter correlation lengths of the random coefficient. We will also qualify the conclusions to state that the superiority is demonstrated for the tested regimes and discuss the conditions under which we expect the advantage to persist. These changes will make the scope and limitations of the performance claims explicit. revision: yes

Circularity Check

0 steps flagged

Direct numerical comparison of distinct methods with no self-referential derivation

full rationale

The paper presents a stochastic Galerkin discretization for random parabolic PDEs, pairs it with a shared GMRES-GMG algebraic solver, and performs head-to-head numerical experiments against a Monte-Carlo approach built on the identical space-time finite-element and solver infrastructure. No central quantity is obtained by fitting a parameter to a subset of the reported data and then re-presenting that fit as a prediction; no uniqueness theorem or ansatz is imported via self-citation to force the formulation; and the observed performance differences are reported as empirical outcomes for the chosen test problems rather than derived by algebraic reduction from the inputs. The derivation chain therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The work rests on standard finite-element convergence theory for parabolic problems and on the well-posedness of stochastic Galerkin projections; no new free parameters are fitted to data and no new physical entities are postulated.

axioms (2)
  • standard math Standard a priori error estimates hold for the space-time finite element discretization of parabolic equations
    Invoked to interpret observed convergence rates in the numerical experiments.
  • domain assumption The stochastic Galerkin projection onto a finite-dimensional polynomial chaos space yields a well-posed coupled system
    Required for the algebraic system to be solvable by the described GMRES-GMG iteration.

pith-pipeline@v0.9.0 · 5736 in / 1482 out tokens · 42643 ms · 2026-05-21T02:02:25.715083+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

  1. [1]

    T. J. Sullivan,Introduction to Uncertainty Quantification, Texts in Applied Mathematics, Vol. 63, Springer, Cham, 2015. doi:10.1007/978-3-319-23395-6

    work page doi:10.1007/978-3-319-23395-6 2015

  2. [2]

    R. E. Caflisch,Monte Carlo and quasi-Monte Carlo methods, Acta Numerica7, 1–49 (1998). doi:10.1017/S0962492900002804

    work page doi:10.1017/s0962492900002804 1998

  3. [3]

    Barth and A

    A. Barth and A. Lang,Multilevel Monte Carlo method with applications to stochastic partial differential equations, International Journal of Computer Mathematics89(18), 2479–2498 (2012)

    work page 2012

  4. [4]

    T. J. Sullivan,Introduction to Uncertainty Quantification, Texts in Applied Mathematics, Vol. 63, Springer, Cham (2015)

    work page 2015

  5. [5]

    M. B. Giles,Multilevel Monte Carlo methods, Acta Numerica24, 259–328 (2015)

    work page 2015

  6. [6]

    J. F. Kenney and E. S. Keeping,Mathematics of Statistics, Part Two, 2nd ed., D. Van Nostrand, Princeton, NJ (1951)

    work page 1951

  7. [7]

    Bierig and A

    C. Bierig and A. Chernov,Estimation of arbitrary order central statistical moments by the multilevel Monte Carlo method, Stochastics and Partial Differential Equations: Analysis and Computations4(1), 3–40 (2016)

    work page 2016

  8. [8]

    R. G. Ghanem and P. D. Spanos,Stochastic Finite Elements: A Spectral Approach, Springer, New York, 1991. doi:10.1007/978-1-4612-3094-6

    work page doi:10.1007/978-1-4612-3094-6 1991

  9. [9]

    Xiu and G

    D. Xiu and G. E. Karniadakis,The Wiener–Askey Polynomial Chaos for Stochastic Differential Equations, SIAM Journal on Scientific Computing24(2), 619–644 (2002). doi:10.1137/S1064827501387826

    work page doi:10.1137/s1064827501387826 2002

  10. [10]

    Xiu,Numerical Methods for Stochastic Computations: A Spectral Method Approach, Princeton University Press, Princeton, NJ, 2010

    D. Xiu,Numerical Methods for Stochastic Computations: A Spectral Method Approach, Princeton University Press, Princeton, NJ, 2010

    work page 2010

  11. [11]

    Thomée,Galerkin Finite Element Methods for Parabolic Problems, 2nd ed., Springer, Berlin, 2006

    V. Thomée,Galerkin Finite Element Methods for Parabolic Problems, 2nd ed., Springer, Berlin, 2006. doi:10.1007/3-540-33122-0

    work page doi:10.1007/3-540-33122-0 2006

  12. [12]

    A. K. Aziz and P. Monk,Continuous finite elements in space and time for the heat equation, Mathematics of Computation52(186), 255–274 (1989)

    work page 1989

  13. [13]

    Schötzau and C

    D. Schötzau and C. Schwab,Time discretization of parabolic problems by the hp-version of the discon- tinuous Galerkin finite element method, SIAM Journal on Numerical Analysis38(3), 837–875 (2000). doi:10.1137/S0036142999352394

    work page doi:10.1137/s0036142999352394 2000

  14. [14]

    Bause and U

    M. Bause and U. Köcher,Variational time discretization for mixed finite element approximations of non- stationary diffusion problems, Journal of Computational and Applied Mathematics289, 208–224 (2015). doi:10.1016/j.cam.2015.02.015

    work page doi:10.1016/j.cam.2015.02.015 2015

  15. [15]

    Bause, F

    M. Bause, F. A. Radu, and U. Köcher,Error analysis for discretizations of parabolic problems using continuous finite elements in time and mixed finite elements in space, Numerische Mathematik137(4), 773–818 (2017). doi:10.1007/s00211-017-0894-6

    work page doi:10.1007/s00211-017-0894-6 2017

  16. [16]

    Nobile and R

    F. Nobile and R. Tempone,Analysis and implementation issues for the numerical approximation of parabolic equations with random coefficients, International Journal for Numerical Methods in Engineering80(6–7), 979– 1006 (2009). doi:10.1002/nme.2656

    work page doi:10.1002/nme.2656 2009

  17. [17]

    Cohen, R

    A. Cohen, R. DeVore, and C. Schwab,Convergence rates of bestN-term Galerkin approximations for a class of elliptic sPDEs, Foundations of Computational Mathematics10(6), 615–646 (2010). doi:10.1007/s10208-010-9072- 2. 27

    work page doi:10.1007/s10208-010-9072- 2010

  18. [18]

    Schwab and C

    C. Schwab and C. J. Gittelson,Sparse tensor discretizations of high-dimensional parametric and stochastic PDEs, Acta Numerica20, 291–467 (2011). doi:10.1017/S0962492911000055

    work page doi:10.1017/s0962492911000055 2011

  19. [19]

    V. H. Hoang and C. Schwab,Sparse Tensor Galerkin Discretization of Parametric and Random Parabolic PDEs— Analytic Regularity and Generalized Polynomial Chaos Approximation, SIAM Journal on Mathematical Analysis 45(5), 3050–3083 (2013). doi:10.1137/100793682

    work page doi:10.1137/100793682 2013

  20. [20]

    Rosseel and S

    E. Rosseel and S. Vandewalle,Iterative Solvers for the Stochastic Finite Element Method, SIAM Journal on Scientific Computing32(1), 372–397 (2010). doi:10.1137/080727026

    work page doi:10.1137/080727026 2010

  21. [21]

    Ullmann,A Kronecker Product Preconditioner for Stochastic Galerkin Finite Element Discretizations, SIAM Journal on Scientific Computing32(2), 923–946 (2010)

    E. Ullmann,A Kronecker Product Preconditioner for Stochastic Galerkin Finite Element Discretizations, SIAM Journal on Scientific Computing32(2), 923–946 (2010). doi:10.1137/080742853

    work page doi:10.1137/080742853 2010

  22. [22]

    Bespalov, D

    A. Bespalov, D. Loghin, and R. Youngnoi,Truncation Preconditioners for Stochastic Galerkin Finite Element Discretizations,SIAM Journal on Scientific Computing43(5), S92–S116 (2021). doi:10.1137/20M1345645

    work page doi:10.1137/20m1345645 2021

  23. [23]

    P. J. Roache,Verification and Validation in Computational Science and Engineering, Hermosa Publishers, Albu- querque, NM, 1998

    work page 1998

  24. [24]

    W. L. Oberkampf and T. G. Trucano,Verification and validation in computational fluid dynamics, Progress in Aerospace Sciences38(3), 209–272 (2002). doi:10.1016/S0376-0421(02)00005-2

    work page doi:10.1016/s0376-0421(02)00005-2 2002

  25. [25]

    Margenberg, M

    N. Margenberg, M. Bause, P. Munch,Anhpmultigrid approach for tensor-product space-time finite element dis- cretizations of the Stokes equations, SIAMJ.Sci.Comput.,47(2025), pp.B1503-B1529; doi: 10.1137/25M1734142

    work page doi:10.1137/25m1734142 2025

  26. [26]

    Margenberg, P

    N. Margenberg, P. Munch,A Space-Time Multigrid Method for Space-Time Finite Element Discretizations of Parabolic and Hyperbolic PDEs, preprint arXiv:2408.04372

    work page arXiv

  27. [27]

    Margenberg and M

    N. Margenberg and M. Bause,A Monolithic hp Space-Time Multigrid Preconditioned Newton-Krylov Solver for Space-Time FEM applied to the Incompressible Navier-Stokes Equations, preprint arXiv:2602.13841, 2026

    work page arXiv 2026

  28. [28]

    Dawor, M

    M. Dawor, M. Bause, and N. Margenberg,Matrix-free multigrid preconditioning for stochastic Galerkin space-time finite element discretizations of random parabolic problems, in preparation

    work page

  29. [29]

    Arndt, W

    D. Arndt, W. Bangerth, M. Bergbauer, B. Blais, M. Fehling, R. Gassmöller, T. Heister, L. Heltai, M. Kronbichler, M. Maier, P. Munch, S. Scheuerman, B. Turcksin, S. Uzunbajakau, D. Wells, and M. Wichrowski,The deal.II library, Version 9.7, Journal of Numerical Mathematics33(4), 403–415 (2025). doi:10.1515/jnma-2025-0115

    work page doi:10.1515/jnma-2025-0115 2025

  30. [30]

    Kronbichler and K

    M. Kronbichler and K. Kormann,A generic interface for parallel cell-based finite element operator application, Computers & Fluids63, 135–147 (2012). doi:10.1016/j.compfluid.2012.04.012. 28

    work page doi:10.1016/j.compfluid.2012.04.012 2012