pith. sign in

arxiv: 2509.02617 · v2 · submitted 2025-09-01 · 📊 stat.ML · cs.LG· stat.CO

Gaussian process surrogate with physical law-corrected prior for multi-coupled PDEs defined on irregular geometry

Pucheng Tang , Hongqiao Wang , Wenzhou Lin , Qian Chen , Heng Yong This is my paper

Pith reviewed 2026-05-18 20:23 UTC · model grok-4.3

classification 📊 stat.ML cs.LGstat.CO
keywords Gaussian processsurrogate modelingparametric PDEsphysics-informedirregular geometryproper orthogonal decompositionRBF-FD
0
0 comments X

The pith

Physical law-corrected prior in Gaussian processes enables surrogate modeling of nonlinear multi-coupled PDEs on irregular domains without kernel redesign.

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

The paper develops a Gaussian process surrogate for parametric PDEs by reducing high-dimensional solutions to a low-dimensional modal space via proper orthogonal decomposition. It then constructs a law-corrected prior that directly incorporates the governing physical laws, overcoming the linear operator invariance that limits existing physics-informed GPs. This allows the approach to handle nonlinear and multi-coupled systems on irregular geometries using RBF-FD differentiation matrices that remain independent of the solution fields. A sympathetic reader would care because repeated high-fidelity simulations across parameter spaces become computationally feasible while preserving physical consistency without per-problem kernel redesign.

Core claim

The authors claim that a physical law-corrected prior Gaussian process (LC-prior GP) overcomes the linear operator invariance limitation of existing physics-informed GP methods by embedding the governing physical laws directly into the prior. Combined with POD for dimensionality reduction and RBF-FD for generating training data on irregular domains, this enables accurate and efficient surrogate modeling of nonlinear multi-parameter and multi-coupled PDE systems without requiring kernel redesign for each application.

What carries the argument

The law-corrected prior (LC-prior), which incorporates governing physical laws to adjust the Gaussian process prior and thereby extends applicability to nonlinear and multi-coupled systems.

If this is right

  • Surrogate optimization occurs efficiently in the low-dimensional POD coefficient space rather than full-order space.
  • The framework applies directly to multi-coupled physical variables on different two-dimensional irregular domains.
  • RBF-FD differentiation matrices can be reused across optimization steps without reassembly since they are independent of solution fields.
  • Extensive tests demonstrate higher accuracy and efficiency than baseline approaches for nonlinear multi-parameter systems.

Where Pith is reading between the lines

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

  • Similar prior-correction techniques could be tested on time-dependent or three-dimensional parametric PDEs to assess broader scalability.
  • The method might connect to uncertainty quantification tasks where embedding physics improves reliability of surrogate predictions across parameter ranges.
  • Integration with other low-dimensional representations beyond POD could further reduce costs for very high-dimensional problems.

Load-bearing premise

Governing physical laws can be directly incorporated into the Gaussian process prior in a manner that overcomes the linear operator invariance limitation for nonlinear multi-coupled cases.

What would settle it

Numerical experiments on a nonlinear multi-coupled PDE system defined on an irregular domain where the LC-prior GP maintains low prediction error while standard physics-informed GPs exhibit large errors due to nonlinearity would support the claim; the reverse outcome would falsify it.

Figures

Figures reproduced from arXiv: 2509.02617 by Heng Yong, Hongqiao Wang, Pucheng Tang, Qian Chen, Wenzhou Lin.

Figure 1
Figure 1. Figure 1: A Schematic of LC-prior GP for differential equations. projection coefficients of the RB model. Song et al. [29] proposed a model reduction method in which the Koopman operator is extracted from the training set via dynamic mode decom￾position (DMD), and the features of a subset of training samples are linearly combined using a nearest-neighbor weighting strategy to enable prediction. As a purely data-driv… view at source ↗
Figure 2
Figure 2. Figure 2: The interpolation point set Y and evaluation point set X. Any point x ∈ Ω is uniquely with one stencil Ys by Eq.(3). Then the local interpolation for the solution of the PDE problem evaluated at the point x with local stencil can be written as: u s h (x) = Xn i=1 c s i ϕ(∥x − y s i ∥) + Xm k=1 β s k pk(x) =  ϕ(∥x − y s 1 ∥), . . . , ϕ(∥x − y s n ∥), p1(x), · · · , pm(x)  [PITH_FULL_IMAGE:figures/full_fi… view at source ↗
Figure 3
Figure 3. Figure 3: A basic strategy for gaining θlaw, the blue points are θobs for training fk(·), and orange points are θlaw for learning physical law correction function ωk(·). global DE constraints as prior knowledge, we enhance the GPR model performance, improv￾ing its ability to predict. This approach is independent of the data-driven nature of the GPR model and does not require additional training data from expensive n… view at source ↗
Figure 4
Figure 4. Figure 4: Relative errors between the mean solutions obtained by the RBF-FD and the LC-prior GP with [PITH_FULL_IMAGE:figures/full_fig_p023_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The first column presents the mean solutions of the RBF-FD method at [PITH_FULL_IMAGE:figures/full_fig_p024_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The first column presents the mean solutions of the RBF-FD method at [PITH_FULL_IMAGE:figures/full_fig_p025_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Two different strategies to chose the physical law correction points. The blue points are GP surrogate training data θobs and orange points are correction points θlaw 4.2 The KdV Equation The Korteweg–de Vries equation is a nonlinear partial differential equation arising in the study of a number of different physical systems, e.g., water waves, plasma physics, anharmonic lattices, and elastic rods [36]. It… view at source ↗
Figure 8
Figure 8. Figure 8: The prediction of u(x, t; θ ∗ ) for KdV equation by different methods. The top-left figure shows the ground truth, the top-right figure represents the standard GP surrogate, and the second row displays LC-prior GP with different number of the physical law corrected points and evaluate their impact on the accuracy of the surrogate model and the MCMC posterior distribution. Fig.8 shows that, compared with th… view at source ↗
Figure 9
Figure 9. Figure 9: The strategy to chose training set and physics correction set (left) and some testing set for a [PITH_FULL_IMAGE:figures/full_fig_p029_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: The first column presents the mean solutions of the RBF-FD method at [PITH_FULL_IMAGE:figures/full_fig_p030_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: The first column presents one physical law-corrected sample’s solutions of the RBF-FD [PITH_FULL_IMAGE:figures/full_fig_p031_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: The first column presents the mean solutions of the RBF-FD method at [PITH_FULL_IMAGE:figures/full_fig_p032_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: The first column presents the mean solutions of the RBF-FD method at [PITH_FULL_IMAGE:figures/full_fig_p034_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: The second derivatives of u in the x-direction and T=0.1 obtained by the RBF-FD(the first and third). The LC-prior GP results are the second one and the fourth one [PITH_FULL_IMAGE:figures/full_fig_p035_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: The first column presents the mean solutions of the RBF-FD method at [PITH_FULL_IMAGE:figures/full_fig_p035_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: The second derivatives of u in the y-direction and T=0.1 obtained by the RBF-FD(the first and third). The LC-prior GP results are the second one and the fourth one. demonstrating its generalization capability. Table.7 provides a comprehensive error analysis, quantifying the aggregate relative errors of the surrogate model across all temporal discretiza￾tion points within [0, T]. Moreover, the precomputed … view at source ↗
read the original abstract

Parametric partial differential equations (PDEs) serve as fundamental mathematical tools for modeling complex physical phenomena, yet repeated high-fidelity numerical simulations across parameter spaces remain computationally prohibitive. In this work, we propose a physical law-corrected prior Gaussian process (LC-prior GP) for efficient surrogate modeling of parametric PDEs. The proposed method employs proper orthogonal decomposition (POD) to represent high-dimensional discrete solutions in a low-dimensional modal coefficient space, significantly reducing the computational cost of kernel optimization compared with standard GP approaches in full-order spaces. The governing physical laws are further incorporated to construct a law-corrected prior to overcome the limitation of existing physics-informed GP methods that rely on linear operator invariance, which enables applications to nonlinear and multi-coupled PDE systems without kernel redesign. Furthermore, the radial basis function-finite difference (RBF-FD) method is adopted for generating training data, allowing flexible handling of irregular spatial domains. The resulting differentiation matrices are independent of solution fields, enabling efficient optimization in the physical correction stage without repeated assembly. The proposed framework is validated through extensive numerical experiments, including nonlinear multi-parameter systems and scenarios involving multi-coupled physical variables defined on different two-dimensional irregular domains to highlight the accuracy and efficiency compared with baseline approaches.

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

2 major / 2 minor

Summary. The manuscript proposes a physical law-corrected prior Gaussian process (LC-prior GP) surrogate for parametric PDEs on irregular geometries. It reduces dimensionality via proper orthogonal decomposition (POD) on modal coefficients, incorporates governing physical laws into the GP prior to address limitations of linear-operator-invariant physics-informed GPs, and thereby claims to handle nonlinear multi-coupled systems without kernel redesign. Training data are generated with radial basis function-finite difference (RBF-FD) discretizations whose differentiation matrices are independent of the solution field, enabling efficient physical correction. Validation consists of numerical experiments on nonlinear multi-parameter and multi-coupled problems defined on two-dimensional irregular domains.

Significance. If the law-corrected prior can be shown to embed nonlinear residuals into the GP without problem-specific linearization or auxiliary parameters, the framework would offer a practical route to physics-informed surrogates for multi-physics systems on complex domains while retaining the computational advantages of POD and mesh-independent differentiation matrices. This would meaningfully extend existing physics-informed GP literature beyond linear operators.

major comments (2)
  1. [Abstract and method description] Abstract and LC-prior construction: the central claim that the law-corrected prior overcomes linear-operator invariance for nonlinear multi-coupled PDEs without kernel redesign is load-bearing. No equations are supplied showing how the nonlinear residual is applied to the GP mean or covariance (e.g., direct evaluation on POD coefficients versus linearization around the current mean), leaving open whether the correction remains general or reduces to per-problem engineering adjustments.
  2. [Numerical experiments] Numerical experiments section: the abstract states that the framework is validated through extensive experiments and highlights accuracy and efficiency gains, yet no quantitative error metrics, convergence rates, or explicit comparison tables against baselines are referenced. This absence prevents verification that the claimed advantages are realized for the nonlinear cases.
minor comments (2)
  1. [Abstract] The abstract would benefit from a single sentence summarizing the key quantitative improvements (e.g., relative L2 errors or speed-up factors) observed in the reported experiments.
  2. [POD and prior construction] Clarify the precise role of the POD coefficient space in the physical correction step; it is stated to reduce kernel-optimization cost, but the interaction with the law-corrected prior is not fully spelled out.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thorough review and valuable feedback on our manuscript. We address the major comments point by point below, providing clarifications and indicating revisions made to the manuscript.

read point-by-point responses

  1. Referee: [Abstract and method description] Abstract and LC-prior construction: the central claim that the law-corrected prior overcomes linear-operator invariance for nonlinear multi-coupled PDEs without kernel redesign is load-bearing. No equations are supplied showing how the nonlinear residual is applied to the GP mean or covariance (e.g., direct evaluation on POD coefficients versus linearization around the current mean), leaving open whether the correction remains general or reduces to per-problem engineering adjustments.

    Authors: We appreciate the referee highlighting the need for explicit mathematical details on the LC-prior. The construction in Section 3 incorporates the nonlinear residual of the governing PDEs directly into the GP prior mean (and covariance via the associated kernel correction), evaluated on the POD modal coefficients without linearization or auxiliary parameters. This is intended to preserve generality for nonlinear multi-coupled systems. To strengthen the presentation, we have added the explicit equations (now labeled (12)–(15)) detailing the residual application to the mean and covariance functions, along with a brief proof sketch confirming no per-problem engineering is required. revision: yes

  2. Referee: [Numerical experiments] Numerical experiments section: the abstract states that the framework is validated through extensive experiments and highlights accuracy and efficiency gains, yet no quantitative error metrics, convergence rates, or explicit comparison tables against baselines are referenced. This absence prevents verification that the claimed advantages are realized for the nonlinear cases.

    Authors: The numerical experiments in Section 4 report quantitative L2 and maximum errors, convergence rates with respect to training set size, and wall-clock timings, with direct comparisons to standard GP, POD-only, and physics-informed GP baselines; these appear in Tables 1–4 and Figures 5–8 for the nonlinear multi-parameter and multi-coupled test cases. To address the referee’s concern about explicit referencing, we have added summary statements of these metrics and convergence rates to the abstract and expanded the opening paragraph of Section 4 to point readers to the specific tables and figures. revision: yes

Circularity Check

0 steps flagged

Law-corrected prior adds independent physical information; no reduction to inputs by construction

full rationale

The derivation chain begins with POD to project high-dimensional PDE solutions onto a low-dimensional modal coefficient space, followed by construction of a law-corrected prior that folds in the governing equations to address nonlinear multi-coupled cases. This correction step is described as overcoming linear-operator invariance limitations of prior physics-informed GPs, supplying external physical constraints rather than re-fitting or re-deriving the kernel from the same data used for prediction. RBF-FD differentiation matrices are independent of solution fields, enabling separate optimization in the correction stage. No self-definitional steps, fitted-input predictions, or load-bearing self-citations appear in the abstract or described framework; the central claim retains independent content from the physical laws and is validated on external numerical benchmarks for irregular domains.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

Abstract-only review prevents exhaustive identification of fitted parameters or background axioms; the law-corrected prior appears to be the main new element introduced without independent evidence supplied.

invented entities (1)
  • law-corrected prior no independent evidence
    purpose: to enable nonlinear and multi-coupled PDE handling without kernel redesign
    Introduced in the abstract to overcome the linear operator invariance limitation of existing physics-informed GP methods.

pith-pipeline@v0.9.0 · 5762 in / 1166 out tokens · 35397 ms · 2026-05-18T20:23:28.285484+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

36 extracted references · 36 canonical work pages · 1 internal anchor

  1. [1]

    Time Delay ODE/PDE Models: Applications in Biomedical Science and Engineering

    WE Schiesser. Time Delay ODE/PDE Models: Applications in Biomedical Science and Engineering. CRC Press, 2019

    work page 2019

  2. [2]

    Tamm: A new topology-aware mapping method for parallel applications on the tianhe-2a su- percomputer

    Xinhai Chen, Jie Liu, Shengguo Li, Peizhen Xie, Lihua Chi, and Qinglin Wang. Tamm: A new topology-aware mapping method for parallel applications on the tianhe-2a su- percomputer. In Algorithms and Architectures for Parallel Processing: 18th International Conference, ICA3PP 2018, Guangzhou, China, November 15-17, 2018, Proceedings, Part I 18, pages 242–256...

    work page 2018

  3. [3]

    Conservative physics- informed neural networks on discrete domains for conservation laws: Applications to for- ward and inverse problems

    Ameya D Jagtap, Ehsan Kharazmi, and George Em Karniadakis. Conservative physics- informed neural networks on discrete domains for conservation laws: Applications to for- ward and inverse problems. Computer Methods in Applied Mechanics and Engineering, 365:113028, 2020. 37

    work page 2020

  4. [4]

    Physics-informed learning machines for partial differential equations: Gaussian processes versus neural networks

    Guofei Pang and George Em Karniadakis. Physics-informed learning machines for partial differential equations: Gaussian processes versus neural networks. Emerging frontiers in nonlinear science, pages 323–343, 2020

    work page 2020

  5. [5]

    Computational fluid dynamics, volume 206

    John David Anderson and John Wendt. Computational fluid dynamics, volume 206. Springer, 1995

    work page 1995

  6. [6]

    A machine learning framework for data driven acceleration of computations of differential equations

    Siddhartha Mishra. A machine learning framework for data driven acceleration of com- putations of differential equations. arXiv preprint arXiv:1807.09519, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  7. [7]

    Kernel methods are competitive for operator learning

    Pau Batlle, Matthieu Darcy, Bamdad Hosseini, and Houman Owhadi. Kernel methods are competitive for operator learning. Journal of Computational Physics, 496:112549, 2024

    work page 2024

  8. [8]

    Operator learning with gaussian processes

    Carlos Mora, Amin Yousefpour, Shirin Hosseinmardi, Houman Owhadi, and Ramin Bostanabad. Operator learning with gaussian processes. Computer Methods in Applied Mechanics and Engineering, 434:117581, 2025

    work page 2025

  9. [9]

    Data-driven identi- fication of parametric partial differential equations

    Samuel Rudy, Alessandro Alla, Steven L Brunton, and J Nathan Kutz. Data-driven identi- fication of parametric partial differential equations. SIAM Journal on Applied Dynamical Systems, 18(2):643–660, 2019

    work page 2019

  10. [10]

    Wavelet neural operator for solving parametric partial differential equations in computational mechanics problems

    Tapas Tripura and Souvik Chakraborty. Wavelet neural operator for solving parametric partial differential equations in computational mechanics problems. Computer Methods in Applied Mechanics and Engineering, 404:115783, 2023

    work page 2023

  11. [11]

    Learning the solution operator of para- metric partial differential equations with physics-informed deeponets

    Sifan Wang, Hanwen Wang, and Paris Perdikaris. Learning the solution operator of para- metric partial differential equations with physics-informed deeponets. Science advances, 7(40):eabi8605, 2021

    work page 2021

  12. [12]

    Bayesian calibration of computer models

    Marc C Kennedy and Anthony O’Hagan. Bayesian calibration of computer models. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 63(3):425– 464, 2001

    work page 2001

  13. [13]

    An improved data-free surrogate model for solving partial di fferential equations using deep neural networks

    Xinhai Chen, Rongliang Chen, Qian Wan, Rui Xu, and Jie Liu. An improved data-free surrogate model for solving partial di fferential equations using deep neural networks. Scientific reports, 11(1):19507, 2021

    work page 2021

  14. [14]

    Surrogate modeling of advanced computer simulations using deep gaussian processes

    Majdi I Radaideh and Tomasz Kozlowski. Surrogate modeling of advanced computer simulations using deep gaussian processes. Reliability Engineering & System Safety, 195:106731, 2020

    work page 2020

  15. [15]

    Physics-enhanced deep surrogates for partial differential equations

    Rapha ¨el Pestourie, Youssef Mroueh, Chris Rackauckas, Payel Das, and Steven G John- son. Physics-enhanced deep surrogates for partial differential equations. Nature Machine Intelligence, 5(12):1458–1465, 2023. 38

    work page 2023

  16. [16]

    Surrogate modeling for bayesian inverse problems based on physics-informed neural networks

    Yongchao Li, Yanyan Wang, and Liang Yan. Surrogate modeling for bayesian inverse problems based on physics-informed neural networks. Journal of Computational Physics, 475:111841, 2023

    work page 2023

  17. [17]

    Reduced-order modeling: new ap- proaches for computational physics

    David J Lucia, Philip S Beran, and Walter A Silva. Reduced-order modeling: new ap- proaches for computational physics. Progress in aerospace sciences, 40(1-2):51–117, 2004

    work page 2004

  18. [18]

    Multilevel model reduction for uncertainty quantification in computational structural dynamics

    Olivier Ezvan, Anas Batou, Christian Soize, and Laurent Gagliardini. Multilevel model reduction for uncertainty quantification in computational structural dynamics. Computational Mechanics, 59(2):219–246, 2017

    work page 2017

  19. [19]

    Adaptive symplectic model order reduction of parametric particle-based vlasov–poisson equation

    Jan Hesthaven, Cecilia Pagliantini, and Nicol `o Ripamonti. Adaptive symplectic model order reduction of parametric particle-based vlasov–poisson equation. Mathematics of Computation, 93(347):1153–1202, 2024

    work page 2024

  20. [20]

    A graph convolutional autoencoder approach to model order reduction for parametrized pdes

    Federico Pichi, Beatriz Moya, and Jan S Hesthaven. A graph convolutional autoencoder approach to model order reduction for parametrized pdes. Journal of Computational Physics, 501:112762, 2024

    work page 2024

  21. [21]

    Gappy spectral proper orthogonal decomposition

    Akhil Nekkanti and Oliver T Schmidt. Gappy spectral proper orthogonal decomposition. Journal of Computational Physics, 478:111950, 2023

    work page 2023

  22. [22]

    Pod-dl-rom: Enhancing deep learning-based re- duced order models for nonlinear parametrized pdes by proper orthogonal decomposition

    Stefania Fresca and Andrea Manzoni. Pod-dl-rom: Enhancing deep learning-based re- duced order models for nonlinear parametrized pdes by proper orthogonal decomposition. Computer Methods in Applied Mechanics and Engineering, 388:114181, 2022

    work page 2022

  23. [23]

    A basis for bounding the errors of proper generalised decom- position solutions in solid mechanics

    JP Moitinho de Almeida. A basis for bounding the errors of proper generalised decom- position solutions in solid mechanics. International Journal for Numerical Methods in Engineering, 94(10):961–984, 2013

    work page 2013

  24. [24]

    An overview of the proper generalized decomposition with applications in computational rheology

    Francisco Chinesta, Amine Ammar, Adrien Leygue, and Roland Keunings. An overview of the proper generalized decomposition with applications in computational rheology. Journal of Non-Newtonian Fluid Mechanics, 166(11):578–592, 2011

    work page 2011

  25. [25]

    Principal component analysis in linear systems: Controllability, observ- ability, and model reduction

    Bruce Moore. Principal component analysis in linear systems: Controllability, observ- ability, and model reduction. IEEE transactions on automatic control, 26(1):17–32, 1981

    work page 1981

  26. [26]

    Interpolatory projection methods for parameterized model reduction

    Ulrike Baur, Christopher Beattie, Peter Benner, and Serkan Gugercin. Interpolatory projection methods for parameterized model reduction. SIAM Journal on Scientific Computing, 33(5):2489–2518, 2011

    work page 2011

  27. [27]

    Non-intrusive reduced order modeling of nonlinear problems using neural networks

    Jan S Hesthaven and Stefano Ubbiali. Non-intrusive reduced order modeling of nonlinear problems using neural networks. Journal of Computational Physics, 363:55–78, 2018. 39

    work page 2018

  28. [28]

    Reduced-order model with an artificial neural network for aerostructural design optimization

    Kyung Hyun Park, Sang Ook Jun, Sung Min Baek, Maeng Hyo Cho, Kwan Jung Yee, and Dong Ho Lee. Reduced-order model with an artificial neural network for aerostructural design optimization. Journal of Aircraft, 50(4):1106–1116, 2013

    work page 2013

  29. [29]

    A model reduction method for parametric dynamical systems defined on complex geometries

    Huailing Song, Yuming Ba, Dongqin Chen, and Qiuqi Li. A model reduction method for parametric dynamical systems defined on complex geometries. Journal of Computational Physics, 506:112923, 2024

    work page 2024

  30. [30]

    Rbf-fd formu- las and convergence properties

    Victor Bayona, Miguel Moscoso, Manuel Carretero, and Manuel Kindelan. Rbf-fd formu- las and convergence properties. Journal of Computational Physics, 229(22):8281–8295, 2010

    work page 2010

  31. [31]

    The overlapped radial basis function-finite di fference (rbf-fd) method: A generalization of rbf-fd

    Varun Shankar. The overlapped radial basis function-finite di fference (rbf-fd) method: A generalization of rbf-fd. Journal of Computational Physics, 342:211–228, 2017

    work page 2017

  32. [32]

    On the role of polynomials in rbf-fd approximations: Ii

    Victor Bayona, Natasha Flyer, Bengt Fornberg, and Gregory A Barnett. On the role of polynomials in rbf-fd approximations: Ii. numerical solution of elliptic pdes. Journal of Computational Physics, 332:257–273, 2017

    work page 2017

  33. [33]

    Using compactly supported radial basis functions to solve partial di fferential equations

    Robert Schaback and Holger Wendland. Using compactly supported radial basis functions to solve partial di fferential equations. WIT Transactions on Modelling and Simulation, 23, 2024

    work page 2024

  34. [34]

    The proper orthogonal decomposition in the analysis of turbulent flows

    Gal Berkooz, Philip Holmes, and John L Lumley. The proper orthogonal decomposition in the analysis of turbulent flows. Annual review of fluid mechanics, 25(1):539–575, 1993

    work page 1993

  35. [35]

    Proper orthogonal descriptors for e fficient and accurate interatomic potentials.Journal of Computational Physics, 480:112030, 2023

    Ngoc Cuong Nguyen and Andrew Rohskopf. Proper orthogonal descriptors for e fficient and accurate interatomic potentials.Journal of Computational Physics, 480:112030, 2023

    work page 2023

  36. [36]

    The korteweg–devries equation: a survey of results

    Robert M Miura. The korteweg–devries equation: a survey of results. SIAM review, 18(3):412–459, 1976. 40

    work page 1976