Pith. sign in

REVIEW 3 major objections 6 minor 15 references

Training a physics-informed neural network first on few collocation points, then transferring the learned parameters as the point set grows, cuts training time by an order of magnitude or more on several benchmark PDEs.

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 · deepseek-v4-flash

2026-08-02 01:25 UTC pith:N2QBB3WF

load-bearing objection Useful 2D results, but the paper is unfinished and the transfer mechanism is unproven. the 3 major comments →

arxiv 2607.14665 v1 pith:N2QBB3WF submitted 2026-07-16 math.NA cs.NA

A new strategy for physics-informed neural networks based on hierarchical collocation point refinement

Minjae Choi , Dukhwan Shin , Youngsoo Yang , Eunjung Lee This is my paper
classification math.NA cs.NA MSC 65N3565N5068T07
keywords physics-informed neural networksmultigrid trainingparameter transfercoarse-to-fine learningcollocation pointsHelmholtz equationspectral biasconvection-diffusion-reaction
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.

MPU-PINNs is a training strategy for physics-informed neural networks that starts with a small set of collocation points, trains the network to a residual tolerance, then transfers the trained parameters to a denser set, repeating until the target resolution is reached. The central claim is that coarse-to-fine initialization lets the network capture global solution structure cheaply, so the expensive finest-level optimization needs far fewer epochs than training from random initialization. Across the benchmarks reported — 2D Poisson, convection-diffusion-reaction, high-frequency Poisson, and Helmholtz — this yields roughly 10–30x shorter wall-clock training times than conventional PINNs while keeping relative errors comparable, with a scaled loss formulation added for oscillatory cases. The method keeps a single network throughout, so the speedup comes from the training schedule and parameter transfer, not from model size or domain decomposition. If the mechanism holds, any single-network PINN implementation can be accelerated by changing only how collocation points are introduced.

Core claim

On the paper's own terms, the discovery is that most of the optimization work in PINN training can be moved from the finest, most expensive collocation set to coarse sets without sacrificing accuracy. Trained on a coarse set first, the network parameters already approximate the low-frequency global structure of the solution; when those parameters initialize the next level, the method reaches the fine-level stopping tolerance in a handful of epochs, whereas conventional PINNs require tens of thousands. The concrete demonstration is in the epoch tables and elapsed-time comparisons: the finest level consumes 1–85 epochs for MPU-PINNs versus 8,344–100,000 for the baselines. For high-frequency pr

What carries the argument

The central object is the multilevel collocation-point schedule with parameter transfer: a sequence of point sets N_1 < N_2 < ... < N_L ending at the target resolution, with the optimized parameters at level ℓ used to initialize level ℓ+1. The cost identity is Cost_MPU ≈ P Σ_ℓ K_ℓ N_ℓ versus Cost_PINN ≈ K_f N_f P, so savings are realized whenever coarse levels are cheap and the finest level needs far fewer iterations. Two point-selection variants are offered: nested addition, where new points are added to the existing set, and random resampling. The supporting mechanism for oscillatory problems is a rescaled loss that divides the Laplacian term by k^2, making the residual scale comparable to

Load-bearing premise

The whole efficiency argument rests on the Section 3 assertion that parameters trained on a coarse collocation set 'typically provide a favorable initialization' for the next finer level; this premise is not proved, and the text's own missing pieces — no 3D experiment despite the abstract and an empty Section 3.5 with an announced comparison — mean the empirical support is currently limited to the 2D benchmarks actually reported.

What would settle it

Run a single-network PINN on a problem where the coarse grid cannot represent the solution's structure (e.g., a sharp interior layer or a high wavenumber with an initial level too coarse), and measure epochs at the finest level. If the final level still requires as many iterations as a from-scratch run — or if the loss decreases at the same rate regardless of whether the initialization comes from a coarser level — then coarse-to-fine parameter transfer is not the source of the speedup, and the reported gains would instead be attributable to the stopping tolerance or epoch cap.

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

Share X Bluesky LinkedIn Reddit HN

If this is right

  • Across the reported 2D Poisson, convection-diffusion-reaction, high-frequency Poisson, and Helmholtz benchmarks, MPU-PINNs cut elapsed training time to roughly 3–10% of conventional PINN time while keeping relative errors comparable.
  • The finest-level epoch counts fall from thousands or tens of thousands (8,344 for 2D Poisson, 31,848 for convection-diffusion-reaction, 97,730 for Helmholtz) to a handful, often 1–85 epochs.
  • Both nested and random point-selection variants produce the speedup; nested selection tends to be faster, random selection slightly more accurate on the 2D Poisson case.
  • The scaled loss formulation (−(1/k^2)Δu = f̃) reduces relative error from about 2.3e-1 to 2.2e-2 in the high-frequency Poisson test and further cuts training time, showing that the PDE reformulation matters for oscillatory solutions.
  • The framework uses a single network with no subdomain interface conditions, so the acceleration is attributed to the training schedule rather than to model size or domain decomposition.

Where Pith is reading between the lines

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

  • A straightforward extension the paper leaves untested is applying the same schedule to time-dependent and inverse PDEs; the paper's cost model predicts comparable or larger savings there because residual evaluation cost grows with the number of collocation points rather than with model complexity.
  • An ablation that fixes the final loss threshold rather than a per-level tolerance and equalizes total compute would isolate whether the speedup comes from coarse-level initialization quality or from early stopping built into the level scheduler.
  • The method could be combined with residual-based adaptive point selection, adding points where the residual is largest at each level, which the paper does not explore.
  • The abstract promises a 3D Poisson experiment and Section 3.5 announces a comparison with residual-weighted PINNs, but neither appears in the body text; the current numerical evidence is therefore limited to the reported 2D cases.

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

3 major / 6 minor

Summary. The paper introduces MPU-PINNs, a training strategy for physics-informed neural networks that starts with a coarse set of collocation points and progressively refines it, transferring the trained parameters to each finer level. A loss-scaling reformulation is also proposed for high-frequency problems. The authors benchmark against conventional PINNs, SA-PINNs, and XPINNs on 2D Poisson, convection–diffusion–reaction, and Helmholtz problems, reporting roughly 10–30x reductions in wall-clock time with comparable accuracy, and claim extension to 3D Poisson.

Significance. If the reported speedups generalize, MPU-PINNs are a simple, appealing drop-in acceleration for single-network PINN training. The paper provides explicit algorithms (Algorithms 1 and 2), a cost model in Section 3.4, and consistent empirical evidence across four 2D benchmarks, including epoch distributions that make the coarse-to-fine behavior transparent. The scaling strategy for high-frequency problems is a useful contribution. However, the manuscript currently contains unfinished text and overclaims 3D results, and the mechanism of the speedup is not isolated experimentally.

major comments (3)
  1. [Section 3.5] Section 3.5 is an unfinished placeholder. It contains the text 'EXPERIMENTS....', bullet points with 'Use examples (test problems) in PIP-IO', and the speculative sentence 'MPU probably has an advantage...' with no results. This is not a completed section and should be either fully written or removed. Its presence makes the manuscript unsuitable for publication in its current form.
  2. [Abstract and Section 6] The abstract and conclusion state that MPU-PINNs are evaluated on 'two- and three-dimensional Poisson equations', but all numerical examples (4.1, 4.2, 5.1, 5.2) are set in Ω=(0,1)^2. No 3D experiment appears in the paper. The claimed scope overstates the evidence; the authors should either add a 3D Poisson experiment or revise the claims to match the actual experiments.
  3. [Section 3.4 and Tables 3, 5, 6, 9, 10] The central explanatory claim—that parameter transfer from coarse to fine levels is what reduces fine-level iterations—is not isolated. Section 3.4's cost model only shows that coarse epochs are cheaper (O(N_l P)); it does not show that transfer reduces the number of fine-level iterations. The observed decreasing epochs per level could also follow from applying the same stopping tolerance ε to nested point sets, with several fine levels terminating after 1 epoch. A control experiment that re-initializes the network at each level (or otherwise ablates warm-starting) is necessary to support the claimed mechanism. Without this, the speedup could be attributed to the multilevel schedule or the stopping criterion rather than to the transferred parameters.
minor comments (6)
  1. [Section 4.1, text near Table 2] 'The time required ... is nearly 5% shorter' is inconsistent with Table 2, where MPU time is about 3% of the conventional PINN time (51.5 s vs 1650.98 s). Should read 'less than 5% of the elapsed time'.
  2. [Section 3.5] 'PIP-IO' is undefined and appears to be a leftover note.
  3. [Throughout] Typos and wording issues: 'leff' (Section 5), 'convectio–diffusion–reaction' (Section 4 intro), 'The another settings' (Section 5.1), 'Figure 2 exhibits' etc. A careful proofread is needed.
  4. [Algorithms 1 and 2] The symbol L is used both for the level counter and the loss value; this is confusing. Rename the loss, e.g., Loss.
  5. [Example 5.1, Table 6] Baselines are capped at 100,000 epochs while MPU-PINNs are reported with total epochs exceeding 130,000 (e.g., Algorithm 1: 99099+33372.8+1+85). The paper should clarify the per-method stopping criteria and whether the cap applies identically.
  6. [Section 4, experimental protocol] No standard deviations or confidence intervals are reported despite five independent runs; adding these would strengthen the robustness of the accuracy and timing comparisons.

Circularity Check

0 steps flagged

No circular derivation: MPU-PINNs is an empirical training schedule benchmarked against exact solutions and independent baselines; the missing ablation affects support for the mechanism, not circularity.

full rationale

The paper's claimed contribution is an algorithmic training schedule, not a derived identity. The speedups are measured directly against known exact solutions and independent baselines (PINN, SA-PINN, XPINN), so no fitted parameter is renamed as a prediction. The efficiency accounting in Sec. 3.4 (Cost_MPU ≈ P Σ K_l N_l) is an exact cost decomposition; the claim K_L ≪ K_f is an empirical assertion about optimizer behavior, not a consequence of the equation, so there is no self-definitional reduction. The coarse-to-fine initialization benefit is assumed and not isolated by a random-reinitialization ablation, and the same tolerance ε is reused at every level, so the epoch tables do not prove the transfer mechanism; however, this is an experimental-support gap, not circularity. References are external prior work; no load-bearing self-citation or imported uniqueness theorem is present. The scaling reformulations (Eqs. (7), (9)) are algebraically equivalent reweightings of the PDE residual, not a renamed empirical pattern used as evidence. The unfinished draft notes in Sec. 3.5 are speculative, but they do not function as derivation inputs. Therefore no load-bearing step reduces to its own input.

Axiom & Free-Parameter Ledger

6 free parameters · 5 axioms · 0 invented entities

Everything here is an algorithmic or experimental choice. The central efficiency claim depends as much on the stopping rule, initial-level choice, and max-epoch cap as on parameter transfer itself.

free parameters (6)
  • Loss weights λ_r, λ_b = Example 4.1: (1, 1e-2); Example 4.2: (1, 1e-3); Eq. (7): (0.1, 1.0); Eq. (9): (1.0, 0.01)
    Manually set per benchmark; they control the balance between residual and boundary terms and affect both convergence speed and final accuracy.
  • Stopping tolerance ε = 5e-6 (Examples 4.1, 4.2, 5.1), 5e-7 (Example 5.2)
    Used as the stopping criterion in Algorithms 1-2; epoch counts and wall-clock times are measured from hitting it, so it is part of the comparison.
  • Initial refinement level = Level 1 for Example 4.1; Level 4 for Examples 4.2 and 5.1; Level 3 for Example 5.2
    Chosen by hand ('starting from level 4 is more effective'); the reported speedups depend on this choice.
  • Collocation level schedule = N_r: 81, 289, 1089, 4225, 16641, 66049, 263169
    Geometric refinement schedule selected by the authors; no sensitivity study is reported.
  • Scaling weights λ_eq, λ_bc = 0.1/1.0 in Eq. (7); 1.0/0.01 in Eq. (9)
    Introduced for high-frequency examples and tuned per problem; they are part of the proposed scaling strategy.
  • Max epoch cap = 100,000 in Examples 5.1 and 5.2
    Baselines and some MPU levels stop at this cap; the reported elapsed-time comparisons are sensitive to it.
axioms (5)
  • domain assumption The exact solutions used as ground truth are correct and the PDEs are well-posed on the stated domains.
    Relative errors are computed against these analytic solutions; no numerical verification of the exact formulas is provided.
  • domain assumption ADAM with learning rate 1e-3 and the fixed network architectures converges to a low residual on each collocation set within the epoch budget when initialized properly.
    The epoch counts in Tables 3, 5, 6, 9, 10 assume this optimizer behavior; no convergence theorem is given.
  • domain assumption Coarse-level optimized parameters provide a favorable initialization for fine-level loss minimization.
    This is the load-bearing empirical premise in Section 3; the paper gives heuristic discussion but no proof.
  • domain assumption The cost per optimization step scales approximately as O(N_l P), where N_l is the number of collocation points and P is the number of parameters.
    Used in Section 3.4 to justify the speedup analysis; ignores fixed overheads, batching, and optimizer-specific costs.
  • standard math The chosen tanh/sin-activated networks are expressive enough to represent the exact solutions.
    Implicit universal-approximation assumption underlying all PINN experiments.

pith-pipeline@v1.3.0-alltime-deepseek · 11687 in / 14325 out tokens · 136833 ms · 2026-08-02T01:25:52.277872+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of A new strategy for physics-informed neural networks based on hierarchical collocation point refinement." pith.science (2026). https://pith.science/paper/N2QBB3WF

@misc{pith2026260714665,
  author       = {Pith},
  title        = {Pith review of: A new strategy for physics-informed neural networks based on hierarchical collocation point refinement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/N2QBB3WF}},
  note         = {Machine review of arXiv:2607.14665}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Physics-informed neural networks (PINNs) offer a flexible framework for solving partial differential equations (PDEs), but training can become computationally expensive when a large number of collocation points are required to accurately enforce the governing equations. To alleviate this cost, we introduce multigrid-based parameter-updated PINNs (MPU-PINNs), a coarse-to-fine training strategy that progressively increases the number of training points throughout the learning process. The proposed approach begins by training a neural network on a coarse set of collocation points and then transfers the learned parameters to successively finer levels. This initialization strategy enables the network to capture the global features of the solution at a relatively low computational cost before refining local details with additional training points. To further improve performance for high-frequency problems, we incorporate a scaling technique that mitigates the effects of spectral bias during training. We evaluate MPU-PINNs on several benchmark PDEs, including two- and three-dimensional Poisson equations, a convection-diffusion-reaction equation, and the Helmholtz equation. Numerical experiments indicate that MPU-PINNs greatly reduce training time while achieving accuracy comparable to that of conventional PINNs and other representative variants such as SA-PINNs and XPINNs. The results further suggest that the proposed coarse-to-fine learning strategy substantially decreases the optimization effort required at finer levels. Overall, MPU-PINNs provide an efficient single-network training framework that enhances the computational efficiency and scalability of PINNs for a broad range of PDE problems.

Figures

Figures reproduced from arXiv: 2607.14665 by Dukhwan Shin, Eunjung Lee, Minjae Choi, Youngsoo Yang.

Figure 1
Figure 1. Figure 1: Schematic illustration of the MPU-PINNs procedure, where the parameters learned at a coarse level are used to initialize the training at the next finer level. 3.2. Coarse to fine: nested point selection In the first step of the proposed algorithm, we begin by selecting a small, random set of collocation points within the computational domain Ω. Using PINN, we train the model to satisfy the given PDE (1). O… view at source ↗
Figure 2
Figure 2. Figure 2: Exact and predicted solutions for Example 4.1: (a) exact solution (b) PINNs (c) SA-PINNs (d) MPU-PINN Algorithm 1 (e) MPU-PINN Algorithm 2 with 𝑁𝑟 = 4225, rather than from the level 1. This is because the exact solution contains a sharp internal transition near the center of domain, and excessively coarse training points, like level 1, may fail to represent this localized structure. Therefore, we think a s… view at source ↗
Figure 3
Figure 3. Figure 3: Exact and predicted solutions for Example 4.2: (a) exact solution (b) PINNs (c) SA-PINNs (d) MPU-PINN Algorithm 1 (e) MPU-PINN Algorithm 2 [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Exact and predicted solutions for Example 5.1: (a) exact solution (b) PINNs (c) SA-PINNs (d) XPINNs (e) MPU-PINN Algorithm 1 (f) MPU-PINN Algorithm 2 [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Exact and predicted solutions for Example 5.2: (a) exact solution (b) PINNs (c) SA-PINNs (d) XPINNs (e) MPU-PINN Algorithm 1 (f) MPU-PINN Algorithm 2 [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

15 extracted references · 1 linked inside Pith

  1. [1]

    Journal of Computational Physics , volume =

    Raissi, Maziar and Perdikaris, Paris and Karniadakis, George Em , title =. Journal of Computational Physics , volume =. 2019 , doi =

    2019

  2. [2]

    Computer Methods in Applied Mechanics and Engineering , volume =

    Wang, Sifan and Wang, Hanwen and Perdikaris, Paris , title =. Computer Methods in Applied Mechanics and Engineering , volume =. 2021 , doi =

    2021

  3. [3]

    and Gholami, Amir and Zhe, Shandian and Kirby, Robert and Mahoney, Michael W

    Krishnapriyan, Aditi S. and Gholami, Amir and Zhe, Shandian and Kirby, Robert and Mahoney, Michael W. , title =. Advances in Neural Information Processing Systems , volume =. 2021 , url =

    2021

  4. [4]

    Journal of Computational Physics , volume =

    Wang, Sifan and Yu, Xinling and Perdikaris, Paris , title =. Journal of Computational Physics , volume =. 2022 , doi =

    2022

  5. [5]

    Computer Methods in Applied Mechanics and Engineering , volume =

    Wang, Sifan and Sankaran, Shyam and Perdikaris, Paris , title =. Computer Methods in Applied Mechanics and Engineering , volume =. 2024 , doi =

    2024

  6. [6]

    Aditya , title =

    Zhao, Zhiyuan and Ding, Xueying and Prakash, B. Aditya , title =. International Conference on Learning Representations , year =. 2307.11833 , archivePrefix =

    Pith/arXiv arXiv

  7. [7]

    and Karniadakis, George Em , title =

    Jagtap, Ameya D. and Karniadakis, George Em , title =. Communications in Computational Physics , volume =. 2020 , doi =

    2020

  8. [8]

    and Kharazmi, Ehsan and Karniadakis, George Em , title =

    Jagtap, Ameya D. and Kharazmi, Ehsan and Karniadakis, George Em , title =. Computer Methods in Applied Mechanics and Engineering , volume =. 2020 , doi =

    2020

  9. [9]

    and Karniadakis, George Em , title =

    Shukla, Khemraj and Jagtap, Ameya D. and Karniadakis, George Em , title =. Journal of Computational Physics , volume =. 2021 , doi =

    2021

  10. [10]

    Communications in Computational Physics , volume =

    Shin, Yeonjong and Darbon, Jerome and Karniadakis, George Em , title =. Communications in Computational Physics , volume =. 2020 , doi =

    2020

  11. [11]

    SIAM Review , volume =

    Lu, Lu and Meng, Xuhui and Mao, Zhiping and Karniadakis, George Em , title =. SIAM Review , volume =. 2021 , doi =

    2021

  12. [12]

    Computer Methods in Applied Mechanics and Engineering , volume =

    Wu, Chenxi and Zhu, Min and Tan, Qinyang and Kartha, Yadhu and Lu, Lu , title =. Computer Methods in Applied Mechanics and Engineering , volume =. 2023 , doi =

    2023

  13. [13]

    SIAM Journal on Scientific Computing , volume =

    Gao, Zhiwei and Yan, Liang and Zhou, Tao , title =. SIAM Journal on Scientific Computing , volume =. 2023 , doi =

    2023

  14. [14]

    Journal of Computational Physics , volume =

    Tang, Kejun and Wan, Xiaoliang and Yang, Chao , title =. Journal of Computational Physics , volume =. 2023 , doi =

    2023

  15. [15]

    and Braga-Neto, Ulisses M

    McClenny, Levi D. and Braga-Neto, Ulisses M. , title =. Journal of Computational Physics , volume =. 2023 , doi =

    2023