Pith. sign in

REVIEW 2 major objections 3 minor 1 cited by

Randomized quasi-Monte Carlo applied to walk on spheres yields variance decay slightly faster than Monte Carlo rates.

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

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

T0 review · grok-4.3

2026-07-04 01:12 UTC pith:YWQZ4FAZ

load-bearing objection RQMC on walk-on-spheres yields modest variance cuts in examples but theory covers only d=2 harmonic functions. the 2 major comments →

arxiv 2605.08483 v2 pith:YWQZ4FAZ submitted 2026-05-08 math.NA cs.NAstat.CO

Randomized quasi-Monte Carlo for walk on spheres

Valerie N. P. Ho , Art B. Owen This is my paper
classification math.NA cs.NAstat.CO
keywords randomized quasi-Monte Carlowalk on spheresboundary value problemsDirichlet conditionsvariance reductionharmonic functionsMinkowski contentnumerical integration
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper examines randomized quasi-Monte Carlo sampling inside the walk on spheres algorithm for Dirichlet boundary value problems in R^d. For two-dimensional harmonic functions the integrands become periodic indicator functions over regions in the torus, and the authors supply conditions on the boundary that let prior discrepancy bounds apply. Tests with four RQMC methods on five examples, some in three dimensions and some with nonzero sources, produce a median variance reduction rate slightly better than O(n^{-1.1}), together with observed reduction factors between 1.8 and 10.7 at sample size 2^17. No single RQMC variant consistently outperforms the others.

Core claim

Randomized quasi-Monte Carlo can be used inside walk on spheres algorithms for boundary value problems. The relevant integrands for harmonic functions in d=2 are periodic indicator functions on the torus T^k whose boundaries satisfy a (k-1)-dimensional Minkowski content condition; this permits application of the He and Wang (2015) results. The resulting estimators exhibit sampling variances that decrease with n at rates slightly better than Monte Carlo, and the approach extends directly to d=3 and to problems containing nonzero source terms.

What carries the argument

Walk on spheres integration combined with randomized quasi-Monte Carlo sampling of periodic indicator functions over torus regions whose boundaries possess (k-1)-dimensional Minkowski content.

Load-bearing premise

The boundaries of the integration regions have (k-1)-dimensional Minkowski content so that the He and Wang variance bounds apply.

What would settle it

A new set of examples in which all four RQMC methods produce variance decay no faster than O(n^{-1}) would falsify the reported median rate.

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

If this is right

  • Variance of the RQMC estimators decreases slightly faster than O(n^{-1.1}) in the median over the tested methods and examples.
  • Variance reduction factors between 1.8 and 10.7 are attained at n=2^{17}.
  • The same framework applies to problems in d=3 and to boundary value problems with nonzero source functions.
  • Multiple values of the dimension k appear in the RQMC estimates for a single problem.
  • None of the four tested RQMC constructions dominates the others across the examples.

Where Pith is reading between the lines

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

  • Similar variance gains may appear when RQMC is substituted into other random-walk or sphere-based solvers for elliptic PDEs.
  • The observed lack of a dominant method indicates that parameter choices inside each RQMC construction remain open to further optimization.
  • The Minkowski content condition may hold for a wider class of geometries arising in practical applications than the five examples considered.
  • Rates observed at n=2^{17} could be checked at larger n to see whether the modest improvement over Monte Carlo persists or saturates.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 3 minor

Summary. The manuscript investigates the use of randomized quasi-Monte Carlo (RQMC) within walk-on-spheres algorithms for Dirichlet boundary-value problems in R^d. For harmonic functions in d=2 it derives conditions ensuring the relevant integrands are periodic indicator functions on the torus whose boundaries possess (k-1)-dimensional Minkowski content, thereby permitting application of the He-Wang (2015) variance bounds. It then reports numerical experiments on five examples (some in d=3, some with nonzero source terms) in which four RQMC constructions produce sampling variances that decay at median rates slightly better than O(n^{-1.1}), with variance-reduction factors between 1.8 and 10.7 at n=2^{17}; no single RQMC method is found to dominate the others.

Significance. If the reported empirical rates prove robust, the work would demonstrate a practical route to variance reduction for a widely used class of Monte-Carlo PDE solvers. The provision of concrete variance numbers and observed rates across five distinct examples, including higher-dimensional and inhomogeneous cases, constitutes a tangible contribution to the numerical evidence base. The theoretical reduction to periodic indicators with Minkowski-content boundaries is a clean, parameter-free step that strengthens the d=2 harmonic analysis.

major comments (2)
  1. [Abstract] Abstract and the paragraph following the statement of the He-Wang applicability: the central empirical claim aggregates results over five examples that explicitly include d=3 and nonzero source terms, yet the only rigorous justification supplied is the reduction to periodic indicators with Minkowski content, which is stated to hold for harmonic functions with d=2. No analogous reduction or verification that the walk-on-spheres integrands in the d=3 or source cases satisfy the hypotheses of He and Wang (2015) is provided; consequently the observed rates in those regimes rest on an unverified extension.
  2. [Abstract] The median-rate claim (slightly better than O(n^{-1.1})) is presented as the principal quantitative finding, but the manuscript supplies neither per-example error bars nor a statistical test that the observed exponents differ from -1. Without these, it is impossible to assess whether the reported improvement is distinguishable from Monte-Carlo variability across the five examples.
minor comments (3)
  1. The lattices employed in the four RQMC constructions are not described; explicit specification of the underlying point sets (e.g., rank-1 lattices, digital nets, or scrambled Sobol') is required for reproducibility.
  2. The abstract states that the variance reduction factors range from 1.8 to 10.7 at n=2^{17}; a table or figure that disaggregates these factors by example and by RQMC method would make the comparison transparent.
  3. The manuscript should cite Liu (2025) as indicated in the revision note and clarify any use of AI tools in the generation or verification of the numerical results.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major point below, agreeing to revise the abstract for clarity on the scope of the theory versus empirical results. We also agree to enhance the presentation of the rate estimates with additional statistical detail where possible.

read point-by-point responses

  1. Referee: [Abstract] Abstract and the paragraph following the statement of the He-Wang applicability: the central empirical claim aggregates results over five examples that explicitly include d=3 and nonzero source terms, yet the only rigorous justification supplied is the reduction to periodic indicators with Minkowski content, which is stated to hold for harmonic functions with d=2. No analogous reduction or verification that the walk-on-spheres integrands in the d=3 or source cases satisfy the hypotheses of He and Wang (2015) is provided; consequently the observed rates in those regimes rest on an unverified extension.

    Authors: We agree that the rigorous reduction to periodic indicator functions with (k-1)-dimensional Minkowski content, enabling the He-Wang (2015) bounds, is derived only for harmonic functions in d=2. The numerical results for d=3 cases and problems with nonzero source terms are presented purely as empirical observations of variance decay under RQMC. We will revise the abstract (and the relevant paragraph) to explicitly separate the theoretical justification, which applies to the d=2 harmonic setting, from the aggregate empirical median rate observed across all five examples. This clarification will prevent any implication that the same theoretical guarantees extend to the non-harmonic or higher-dimensional cases. revision: yes

  2. Referee: [Abstract] The median-rate claim (slightly better than O(n^{-1.1})) is presented as the principal quantitative finding, but the manuscript supplies neither per-example error bars nor a statistical test that the observed exponents differ from -1. Without these, it is impossible to assess whether the reported improvement is distinguishable from Monte-Carlo variability across the five examples.

    Authors: The median rate is reported as a descriptive summary statistic aggregating the fitted exponents over the four RQMC methods and five examples. The manuscript does not include per-example error bars on the individual exponents nor a formal statistical test against the Monte Carlo baseline of -1. In revision we will add bootstrap-based confidence intervals for the per-example exponents (computed from the available replicate runs) to quantify variability and allow readers to judge whether the observed improvement is distinguishable from sampling fluctuation. We retain the median as a robust central tendency measure but will present it alongside these intervals. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical rates measured directly from simulations

full rationale

The paper reports measured sampling variances and decay rates from explicit RQMC runs on walk-on-spheres integrands across five examples. The He-Wang (2015) citation is invoked only to establish Minkowski-content conditions for the d=2 harmonic integrands; it is not used to compute or normalize the reported O(n^{-1.1}) rates or variance-reduction factors, which are obtained by direct Monte Carlo estimation on the actual sample paths. No fitted parameter is relabeled as a prediction, no self-citation chain is load-bearing, and the d=3/source results are presented purely as numerical observations without claiming the cited theorem applies to them.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the applicability of He and Wang (2015) variance results once the Minkowski-content condition holds for the periodic indicator integrands arising in the walk-on-spheres formulation.

axioms (1)
  • domain assumption ∂Θ has k-1 dimensional Minkowski content
    Invoked to justify application of He and Wang (2015) results to the periodic indicator functions over Θ in T^k.

pith-pipeline@v0.9.1-grok · 5741 in / 1144 out tokens · 25949 ms · 2026-07-04T01:12:20.343759+00:00 · methodology

0 comments
read the original abstract

We investigate the use of randomized quasi-Monte Carlo (RQMC) in walk on spheres algorithms to solve boundary value problems for functions with Dirichlet boundary conditions in $\mathbb{R}^d$. For harmonic functions with $d=2$, the integrands of interest are periodic indicator functions over regions $\Theta$ in the torus $\mathbb{T}^k$. We give conditions for $\partial\Theta$ to have $k-1$ dimensional Minkowski content which allows us to use results of He and Wang (2015). The RQMC estimates involve multiple values of $k$. We see sampling variances decreasing with the number $n$ of sample points at slightly better than Monte Carlo rates. The median variance rate in $4$ RQMC methods over $5$ worked examples, including some with $d=3$ and some with nonzero source functions, was slightly better than $O(n^{-1.1})$. The variance reduction factors ranged from $1.8$ to $10.7$ at $n=2^{17}$. None of the four RQMC methods dominated the others. Changes: cite Liu (2025), describe the lattices that were used, describe usage of AI

Figures

Figures reproduced from arXiv: 2605.08483 by Art B. Owen, Valerie N. P. Ho.

Figure 1
Figure 1. Figure 1: In this problem u(z) is the temperature of the gasket at z ∈ Ω. The purpose of the head gasket is to avoid any leakage of oil or coolant from the combustion engine into the cylinders where combustion occurs. Leaks are more likely to occur where the gasket is hotter. They are interested in approximating the solution to the BVP with ∆u(z) = 0 for z ∈ Ω ◦ and boundary values u(z) = X r∈{coolant, outer, oil, o… view at source ↗
Figure 1
Figure 1. Figure 1: Cylinder head gasket domain. The domain Ω is the interior of the [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Variance of the standard WOS estimator at [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The curve A is a circular arc. Together with the shown rays it defines 4 open regions of R 2 on which rA is analytic. Ramamurthy (1998) which uses parameterized arcs. We write our real analytic arc as A = {ϕ(t) | 0 ⩽ t ⩽ 1} and we assume that ϕ ′ (t) ̸= 0 at any t ∈ [0, 1]. With this parameterization, we define an internal curve A◦ = ϕ((0, 1)) and endpoints a = ϕ(0) and b = ϕ(1). Then rA(z) = min ra(z), rA… view at source ↗
Figure 4
Figure 4. Figure 4: This figure illustrates a portion of the proof of Theorem 10 for [PITH_FULL_IMAGE:figures/full_fig_p022_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: MSE of the WOS estimator for the unit disk example, starting at [PITH_FULL_IMAGE:figures/full_fig_p025_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: MSE of the standard WOS estimator for the major circular sector [PITH_FULL_IMAGE:figures/full_fig_p026_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Dumbbell-shaped domain with parameters R = 1, L = 1.5 and w = 0.4. The starting point z0 = (0.5, 0) used in our experiments is marked with a red cross. Using MC and RQMC samples, we get the results presented in [PITH_FULL_IMAGE:figures/full_fig_p027_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Variance of the WOS estimator for the dumbbell example, started [PITH_FULL_IMAGE:figures/full_fig_p028_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: MSE of the standard WOS estimator for the unit ball example, for [PITH_FULL_IMAGE:figures/full_fig_p029_9.png] view at source ↗

Review history (2 revisions) →

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. Walk on spheres and Array-RQMC

    math.NA 2026-05 unverdicted novelty 7.0

    Array-RQMC-WOS cuts Monte Carlo variance by 57-2290 times with empirical rates n^{-1.4} to n^{-1.8} and introduces a column-wise mean dimension to explain the gain.

Reference graph

Works this paper leans on

36 extracted references · 36 canonical work pages · cited by 1 Pith paper · 1 internal anchor

  1. [1]

    Basu, S. and S. Prasad (2023). A connection between cut locus, T hom space and M orse-- B ott functions. Algebraic & Geometric Topology\/ 23\/ (9), 4185--4233

    work page 2023

  2. [2]

    Binder, I. and M. Braverman (2012). The rate of convergence of the walk on spheres algorithm. Geometric and Functional Analysis\/ 22\/ (3), 558--587

    work page 2012

  3. [3]

    Choi, H. I., S. W. Choi, and H. P. Moon (1997). Mathematical theory of medial axis transform. Pacific Journal of Mathematics\/ 181\/ (1), 57--88

    work page 1997

  4. [4]

    Choi, S.-C. T., F. J. Hickernell, M. McCourt, J. Rathinavel, and A. G. Sorokin (2026). QMCPy : A Q uasi- M onte C arlo P ython L ibrary

    work page 2026

  5. [5]

    Gkioulekas, B

    Crane, K., I. Gkioulekas, B. Miller, and R. Sawhney (2025). Walk on spheres in one weekend. https://brickisland.net/wos-one-weekend/. Online tutorial, accessed April 2026

    work page 2025

  6. [6]

    Walking on Spheres and Talking to Neighbors: Variance Reduction for Laplace's Equation

    Czekanski, M., B. Faber, M. Fairborn, A. Wright, and D. Bindel (2024). Walking on spheres and talking to neighbors: Variance reduction for Laplace's equation. Technical report, arXiv:2404.17692

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  7. [7]

    Dick, J. and F. Pillichshammer (2010). Digital sequences, discrepancy and quasi- Monte Carlo integration . Cambridge: Cambridge University Press

    work page 2010

  8. [8]

    Dudek, E. and K. Holly (1994). Nonlinear orthogonal projection. Annales Polonici Mathematici\/ 59\/ (1), 1--31

    work page 1994

  9. [9]

    Evans, L. C. and R. F. Gariepy (2015). Measure Theory and Fine Properties of Functions\/ (Revised ed.). Textbooks in Mathematics. Boca Raton, FL: CRC Press

    work page 2015

  10. [10]

    Fang, K.-T. and Y. Wang (1994). Number Theoretic Methods in Statistics . Chapman & Hall

    work page 1994

  11. [11]

    Farouki, R. T. and R. Ramamurthy (1998). Degenerate point/curve and curve/curve bisectors arising in medial axis computations for planar domains with curved boundaries. Computer Aided Geometric Design\/ 15\/ (6), 615--635

    work page 1998

  12. [12]

    Federer, H. (1969). Geometric measure theory . Berlin: Springer-Verlag

    work page 1969

  13. [13]

    Foote, R. L. (1984). Regularity of the distance function. Proceedings of the American Mathematical Society\/ 92\/ (1), 153--155

    work page 1984

  14. [14]

    Guillemin, V. and A. Pollack (1974). Differential topology . Englewood Cliffs, NJ: Prentice-Hall

    work page 1974

  15. [15]

    He, Z. and X. Wang (2015). On the convergence rate of randomized quasi- Monte Carlo for discontinuous functions. SIAM Journal on Numerical Analysis\/ 53\/ (5), 2488--2503

    work page 2015

  16. [16]

    Joe, S. and F. Y. Kuo (2008). Constructing Sobol' sequences with better two-dimensional projections. SIAM Journal on Scientific Computing\/ 30\/ (5), 2635--2654

    work page 2008

  17. [17]

    Krantz, S. G. and H. R. Parks (1999). The geometry of domains in space . New York: Springer Science & Business Media

    work page 1999

  18. [18]

    Lee, J. M. (2013). Introduction to Smooth Manifolds\/ (Second ed.). New York: Springer

    work page 2013

  19. [19]

    Leobacher, G. and A. Steinicke (2021). Existence, uniqueness and regularity of the projection onto differentiable manifolds. Annals of global analysis and geometry\/ 60\/ (3), 559--587

    work page 2021

  20. [20]

    Lundberg, E. and K. Ramachandran (2019). A note on the critical points of the torsion function. Technical report, arxiv:1907.08376v1

    work page arXiv 2019

  21. [21]

    Magnanini, R. and G. Poggesi (2022). The location of hot spots and other extremal points. Mathematische Annalen\/ 384\/ (1), 1--39

    work page 2022

  22. [22]

    W alk O n S pheres

    Mascagni, M. and C.-O. Hwang (2003). -shell error analysis for “ W alk O n S pheres” algorithms. Mathematics and computers in simulation\/ 63\/ (2), 93--104

    work page 2003

  23. [23]

    Karaivanova, and C.-O

    Mascagni, M., A. Karaivanova, and C.-O. Hwang (2004). Quasi- Monte Carlo methods for elliptic BVP s. In Monte Carlo and Quasi-Monte Carlo Methods 2002: Proceedings of a Conference held at the National University of Singapore, Republic of Singapore, November 25--28, 2002 , pp.\ 345--355. Springer

    work page 2004

  24. [24]

    Matou s ek, J. (1998). On the L ^2 --discrepancy for anchored boxes. Journal of Complexity\/ 14 , 527--556

    work page 1998

  25. [25]

    Mattila, P. (1999). Geometry of sets and measures in Euclidean spaces : fractals and rectifiability . Number 44. Cambridge, MA: Cambridge University Press

    work page 1999

  26. [26]

    Muller, M. E. (1956). Some continuous Monte Carlo methods for the Dirichlet problem. The Annals of Mathematical Statistics\/ 27\/ (3), 569--589

    work page 1956

  27. [27]

    Niederreiter, H. (1992). Random Number Generation and Quasi- Monte Carlo Methods . Philadelphia, PA: S.I.A.M

    work page 1992

  28. [28]

    Owen, A. B. (2005). Multidimensional variation for quasi- Monte Carlo . In J. Fan and G. Li (Eds.), International Conference on Statistics in honour of Professor Kai-Tai Fang's 65th birthday

    work page 2005

  29. [29]

    Owen, A. B. (2023). Practical Quasi-Monte Carlo Integration . https://artowen.su.domains/mc/practicalqmc.pdf

    work page 2023

  30. [30]

    Owen, A. B. and Z. Pan (2024). Gain coefficients for scrambled Halton points. SIAM Journal on Numerical Analysis\/ 62\/ (3), 1021--1038

    work page 2024

  31. [31]

    Owen, A. B. and D. Rudolf (2021). A strong law of large numbers for scrambled net integration. SIAM Review\/ 63\/ (2), 360--372

    work page 2021

  32. [32]

    Sabelfeld, K. K. (1991). Monte Carlo methods in boundary value problems . Berlin: Springer-Verlag

    work page 1991

  33. [33]

    Sawhney, R. and K. Crane (2020). Monte Carlo geometry processing: A grid-free approach to PDE -based methods on volumetric domains. ACM Transactions on Graphics\/ 39\/ (4)

    work page 2020

  34. [34]

    Miller, I

    Sawhney, R., B. Miller, I. Gkioulekas, and K. Crane (2023). Walk on stars: A grid-free Monte Carlo method for PDE s with Neumann boundary conditions. Technical report, arXiv:2302.11815

    work page arXiv 2023

  35. [35]

    Sloan, I. H. and S. Joe (1994). Lattice Methods for Multiple Integration . Oxford: Oxford Science Publications

    work page 1994

  36. [36]

    Morrical, S

    Wu, L., N. Morrical, S. P. Bangaru, R. Sawhney, S. Zhao, C. Wyman, R. Ramamoorthi, and A. Lefohn (2025). Unbiased differential visibility using fixed-step walk-on-spherical-caps and closest silhouettes. ACM Transactions on Graphics (TOG)\/ 44\/ (4), 1--16

    work page 2025