pith. sign in

arxiv: 2607.01534 · v1 · pith:MAHDNMGUnew · submitted 2026-07-01 · 🧮 math.NA · cs.NA

A spectral-subspace-augmented POD-Galerkin method for parametrized PDEs with limited snapshot data

Tianhao Hu , Zecheng Gan This is my paper

Pith reviewed 2026-07-03 18:55 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords reduced order modelingPOD-Galerkinparametrized PDEsspectral methodslimited snapshot dataDEIMGalerkin projection
0
0 comments X

The pith

SS-POD augments POD-Galerkin with partitioned spectral subspaces to achieve high accuracy from limited snapshots.

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

The paper introduces the spectral-subspace-augmented POD-Galerkin method (SS-POD) for parametrized PDEs when only a small number of solution snapshots can be computed. It begins with a problem-adapted spectral space, divides it into orthogonal subspaces, and applies POD within each subspace after projecting the snapshots. An energy-balancing rule ensures each local POD problem gets roughly the same amount of snapshot energy. Tests on a Laplace-Beltrami equation on the sphere demonstrate that with five snapshots SS-POD reaches errors orders of magnitude smaller than standard POD while using fewer basis functions than a full spectral approach.

Core claim

By partitioning a spectral approximation space into orthogonal subspaces and performing local POD on the projected snapshot matrices, with partitions chosen so that each subspace receives comparable snapshot energy, SS-POD produces reduced bases that maintain high out-of-sample accuracy for parametrized PDEs even when snapshot data is severely limited.

What carries the argument

The energy-balancing partition of a spectral approximation space into orthogonal subspaces for local POD application.

If this is right

  • SS-POD yields compact reduced bases that generalize better than standard POD when snapshots are few.
  • Accuracy gains appear in both linear and nonlinear parametrized PDEs when paired with DEIM.
  • Computational cost for many-query tasks decreases because fewer high-fidelity solves are needed.
  • The approach suits multiscale problems in energy science where generating many snapshots is impractical.

Where Pith is reading between the lines

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

  • If the spectral space is chosen from a standard basis like spherical harmonics, the method might extend to other geometries with known eigenfunction bases.
  • Energy balancing could be replaced by other criteria such as variance or residual norms in future variants.
  • The local POD problems might be solved in parallel to further speed up basis construction.

Load-bearing premise

The initial spectral approximation space must be problem-adapted enough that its orthogonal subspaces can be ranked by the energy content of the available snapshots.

What would settle it

Running the method on a problem where the chosen spectral space does not align with the solution manifold, then checking if the error improvement over standard POD vanishes.

Figures

Figures reproduced from arXiv: 2607.01534 by Tianhao Hu, Zecheng Gan.

Figure 1
Figure 1. Figure 1: Workflow of the SS-POD methodology. Let Φspec = [e1, . . . , eNmax ] be a collection of Nmax normalized spectral ba￾sis functions. We assume that Nmax is large enough for the truncated spectral space E = span{e1, . . . , eNmax } to approximate the relevant solution states to the desired accuracy. Depending on the geometry and boundary conditions, the basis {ej} may be chosen from Fourier modes, Chebyshev p… view at source ↗
Figure 2
Figure 2. Figure 2: SS-POD basis obtained for the Allen-Cahn equation with three snap [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Poisson equation results: source f (left), reference solution (middle￾left), and error maps for POD (middle-right, Nbasis = 15, ϵr P = 3.70 × 10−4 ) and SS-POD (right, Nbasis = 361, ϵr S = 5.92 × 10−7 ). Nsub is the number of subspaces generated by Algorithm 2. For SS-POD, we use tensor-product Chebyshev functions Ti,j (x, y) = Ti(x)Tj (y). (38) The two-dimensional indices (i, j) are ordered by k = p i 2 +… view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of spectral-Galerkin, POD, and SS-POD results across the [PITH_FULL_IMAGE:figures/full_fig_p018_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Helmholtz benchmarks: reference solution at [PITH_FULL_IMAGE:figures/full_fig_p019_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: 1D Heat equation benchmarks: reference evolution (left) and temporal [PITH_FULL_IMAGE:figures/full_fig_p020_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: 1D Allen-Cahn equation benchmark: reference evolution (left) and error [PITH_FULL_IMAGE:figures/full_fig_p020_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Interfacial coarsening in the Allen-Cahn equation. For the reported [PITH_FULL_IMAGE:figures/full_fig_p022_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Laplace-Beltrami benchmarks on a unit sphere. The top row displays [PITH_FULL_IMAGE:figures/full_fig_p024_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Generalization and convergence analysis for the Laplace-Beltrami [PITH_FULL_IMAGE:figures/full_fig_p025_10.png] view at source ↗
read the original abstract

Parametrized partial differential equations (PDEs) arise in many-query simulation, optimization, control, and uncertainty quantification, where repeated full-order solves restrict the number of high-fidelity snapshots that can be generated. This limitation is particularly pronounced in computational energy science, where multiscale models of porous-media flow, transport, and energy materials often make large snapshot datasets impractical. Proper orthogonal decomposition (POD) constructs compact reduced bases from solution snapshots, but it may exhibit limited out-of-sample predictive capability when the snapshots insufficiently sample the solution manifold. To address this limitation, we propose a spectral-subspace-augmented POD-Galerkin method (SS-POD) tailored to limited-data regimes. SS-POD starts from a problem-adapted spectral approximation space, partitions it into orthogonal subspaces, and performs POD locally on the projected snapshot matrices. An energy-balancing rule determines the spectral partition so that the resulting local POD problems are assigned comparable amounts of snapshot energy. For nonlinear parametrized PDEs, SS-POD is coupled with the discrete empirical interpolation method (DEIM). Numerical experiments show that SS-POD improves out-of-sample accuracy over standard POD-Galerkin while retaining compact reduced bases in limited-snapshot regimes. In particular, for a Laplace-Beltrami problem on the unit sphere with only 5 snapshots, SS-POD achieves a relative error of $3.9*10^{-8}$ using 91 basis functions, whereas the standard POD error saturates at $7.8*10^{-4}$ and the spectral-Galerkin method requires 256 basis functions for comparable accuracy. These results indicate that SS-POD provides an effective strategy for high-fidelity reduced-order modeling from limited snapshot data.

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

3 major / 2 minor

Summary. The manuscript proposes a spectral-subspace-augmented POD-Galerkin (SS-POD) method for reduced-order modeling of parametrized PDEs under limited snapshot data. SS-POD starts from a problem-adapted spectral approximation space, partitions it into orthogonal subspaces using an energy-balancing rule so that local POD problems receive comparable snapshot energy, and applies POD locally on the projected snapshot matrices; for nonlinear problems it is coupled with DEIM. Numerical results on a Laplace-Beltrami problem on the unit sphere with only 5 snapshots report that SS-POD attains a relative error of 3.9×10^{-8} with 91 basis functions, while standard POD saturates at 7.8×10^{-4} and a pure spectral-Galerkin method requires 256 functions for comparable accuracy.

Significance. If the central numerical gains hold under the stated assumptions, the method offers a concrete strategy for improving out-of-sample accuracy in data-scarce regimes that arise in energy-science applications. The explicit comparison against both POD and spectral-Galerkin on the sphere problem, together with the use of an energy-balancing partition, supplies a reproducible demonstration that the augmentation can be effective when a high-quality spectral basis is already available. The absence of an error analysis or a general procedure for constructing the initial spectral space, however, restricts the assessed significance to the specific test case shown.

major comments (3)
  1. [Abstract, §3] Abstract and §3 (method description): the claim that SS-POD is applicable to general parametrized PDEs rests on the existence of a 'problem-adapted spectral approximation space' whose orthogonal subspaces admit meaningful energy ranking. No procedure is given for constructing or selecting this space when the operator lacks an obvious eigenbasis (unlike the spherical-harmonics case for the Laplace-Beltrami operator), so the reported accuracy gain cannot be assumed to transfer.
  2. [Abstract] Abstract: the energy-balancing rule that determines the spectral partition is presented without derivation or convergence analysis. Because the partition directly controls the local POD problems and the final basis size (91 functions in the reported experiment), the lack of justification for the rule is load-bearing for the central accuracy claim.
  3. [Numerical experiments] Numerical experiments (sphere Laplace-Beltrami): while the 3.9×10^{-8} error with 5 snapshots is striking, the manuscript provides no comparison against other limited-data techniques (e.g., sparse POD, dictionary learning, or physics-informed ROMs) that also aim to mitigate snapshot scarcity, leaving open whether the gain is specific to the spectral-augmentation strategy.
minor comments (2)
  1. [§3] Notation for the projected snapshot matrices and the energy measure used in the balancing rule should be introduced with explicit definitions before the algorithm is stated.
  2. [Numerical experiments] The manuscript should clarify whether the reported basis-function counts (91 vs. 256) are chosen by the same tolerance or by a fixed dimension; this affects the fairness of the comparison.

Simulated Author's Rebuttal

3 responses · 2 unresolved

We thank the referee for the thoughtful review and constructive comments on our manuscript. We address each major comment below and indicate where revisions will be made.

read point-by-point responses

  1. Referee: [Abstract, §3] Abstract and §3 (method description): the claim that SS-POD is applicable to general parametrized PDEs rests on the existence of a 'problem-adapted spectral approximation space' whose orthogonal subspaces admit meaningful energy ranking. No procedure is given for constructing or selecting this space when the operator lacks an obvious eigenbasis (unlike the spherical-harmonics case for the Laplace-Beltrami operator), so the reported accuracy gain cannot be assumed to transfer.

    Authors: We agree that the method presupposes the availability of a suitable problem-adapted spectral space. The manuscript demonstrates the augmentation strategy in cases where such a space exists (e.g., spherical harmonics for the Laplace-Beltrami operator), without providing a general construction procedure for arbitrary operators. We will revise the abstract and §3 to clarify the scope and assumptions regarding the initial spectral space. revision: partial

  2. Referee: [Abstract] Abstract: the energy-balancing rule that determines the spectral partition is presented without derivation or convergence analysis. Because the partition directly controls the local POD problems and the final basis size (91 functions in the reported experiment), the lack of justification for the rule is load-bearing for the central accuracy claim.

    Authors: The energy-balancing rule is a practical heuristic intended to assign comparable snapshot energy to each local POD subproblem. While we do not supply a formal derivation or convergence analysis, the rule is motivated by numerical stability considerations and is supported by the reported experiments. We will expand the motivation and description of the rule in the revised method section. revision: partial

  3. Referee: [Numerical experiments] Numerical experiments (sphere Laplace-Beltrami): while the 3.9×10^{-8} error with 5 snapshots is striking, the manuscript provides no comparison against other limited-data techniques (e.g., sparse POD, dictionary learning, or physics-informed ROMs) that also aim to mitigate snapshot scarcity, leaving open whether the gain is specific to the spectral-augmentation strategy.

    Authors: The experiments compare SS-POD directly to standard POD-Galerkin and pure spectral-Galerkin to isolate the contribution of the spectral-subspace augmentation. Broader comparisons to other limited-data methods lie outside the intended scope of the paper. The existing results suffice to support the claims regarding improvement over the two baseline approaches in the tested regime. revision: no

standing simulated objections not resolved
  • A general procedure for constructing the problem-adapted spectral approximation space for arbitrary parametrized PDEs
  • An error analysis or convergence theory for the SS-POD method

Circularity Check

0 steps flagged

No circularity; constructive procedure validated on external benchmarks

full rationale

The paper defines SS-POD as an explicit algorithmic procedure (start from a stated problem-adapted spectral space, apply an energy-balancing partition rule, perform local POD on projected snapshots, couple with DEIM for nonlinear cases). The headline numerical claim is an empirical measurement on the unit-sphere Laplace-Beltrami problem (5 snapshots, 91 basis functions, 3.9e-8 error) compared against standard POD and full spectral-Galerkin on the same external test case. No equation reduces a claimed prediction to a fitted input by construction, no load-bearing step rests on a self-citation, and the precondition of a suitable initial spectral space is declared up-front rather than derived from the method itself. The derivation chain is therefore self-contained against the reported benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The method rests on the existence of a useful spectral approximation space that can be orthogonally partitioned and on the validity of the energy-balancing rule; these are domain assumptions rather than derived results. No free parameters or invented entities are named in the abstract.

axioms (1)
  • domain assumption A problem-adapted spectral approximation space exists whose orthogonal subspaces can be ranked by the amount of snapshot energy they carry.
    The partition step and the subsequent local POD both presuppose that such a space has already been constructed and that its energy distribution is meaningful.

pith-pipeline@v0.9.1-grok · 5851 in / 1560 out tokens · 18839 ms · 2026-07-03T18:55:32.977261+00:00 · methodology

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

72 extracted references · 48 canonical work pages · 2 internal anchors

  1. [1]

    WIREs Comput

    Abdi, H., Williams, L.: Principal component analysis. WIREs Comput. Stat.2(4), 433–459 (2010). DOI https://doi.org/10.1002/wics.101

    work page doi:10.1002/wics.101 2010

  2. [2]

    Barrault, M., Maday, Y., Nguyen, N., Patera, A.: An ‘empirical interpolation’ method: Application to efficient reduced-basis discretization of partial differential equations. C. R. Math.339(9), 667–672 (2004). DOI https://doi.org/10.1016/j.crma.2004.08.006

    work page doi:10.1016/j.crma.2004.08.006 2004

  3. [3]

    Binev, P., Cohen, A., Dahmen, W., DeVore, R., Petrova, G., Wojtaszczyk, P.: Con- vergence rates for greedy algorithms in reduced basis methods. SIAM J. Math. Anal. 43(3), 1457–1472 (2011). DOI https://doi.org/10.1137/100795772

    work page doi:10.1137/100795772 2011

  4. [4]

    ESAIM: M2AN 46(3), 595–603 (2012)

    Buffa, A., Maday, Y., Patera, A., Prud’homme, C., Turinici, G.: A priori convergence of the greedy algorithm for the parametrized reduced basis method. ESAIM: M2AN 46(3), 595–603 (2012). DOI https://doi.org/10.1051/m2an/2011056

    work page doi:10.1051/m2an/2011056 2012

  5. [5]

    Springer Science & Business Media, Berlin (2007)

    Canuto, C., Hussaini, M., Quarteroni, A., Zang, T.: Spectral Methods: Evolution to Complex Geometries and Applications to Fluid Dynamics. Springer Science & Business Media, Berlin (2007). DOI https://doi.org/10.1007/978-3-540-30728-0

    work page doi:10.1007/978-3-540-30728-0 2007

  6. [6]

    In: Proc

    Chaturantabut, S., Sorensen, D.: Discrete empirical interpolation for nonlinear model reduction. In: Proc. 48th IEEE Conf. Decis. Control (CDC), pp. 4316–4321 (2009). DOI https://doi.org/10.1109/CDC.2009.5400045

    work page doi:10.1109/cdc.2009.5400045 2009

  7. [7]

    Chaturantabut, S., Sorensen, D.: Nonlinear model reduction via discrete empirical in- terpolation. SIAM J. Sci. Comput.32(5), 2737–2764 (2010). DOI https://doi.org/10. 1137/090766498

    2010

  8. [8]

    Chaturantabut, S., Sorensen, D.: A state space error estimate for POD-DEIM nonlinear model reduction. SIAM J. Numer. Anal.50(1), 46–63 (2012). DOI https://doi.org/10. 1137/110822724

    2012

  9. [9]

    Chen, J., Chi, X., E, W., Yang, Z.: Bridging traditional and machine learning-based algorithms for solving PDEs: The random feature method. J. Mach. Learn.1, 268–298 (2022). DOI https://doi.org/10.4208/jml.220726

    work page doi:10.4208/jml.220726 2022

  10. [10]

    Chen, Y., Ji, L., Narayan, A., Xu, Z.: L1-based reduced over collocation and hyper reduction for steady state and time-dependent nonlinear equations. J. Sci. Comput.87, 10 (2021). DOI https://doi.org/10.1007/s10915-021-01416-z

    work page doi:10.1007/s10915-021-01416-z 2021

  11. [11]

    Chen, Y., Ji, Y., Narayan, A., Xu, Z.: TGPT-PINN: Nonlinear model reduction with transformed GPT-PINNs. Comput. Methods Appl. Mech. Eng.430, 117198 (2024). DOI https://doi.org/10.1016/j.cma.2024.117198

    work page doi:10.1016/j.cma.2024.117198 2024

  12. [12]

    Chen, Y., Koohy, S.: GPT-PINN: Generative pre-trained physics-informed neural net- works toward non-intrusive meta-learning of parametric PDEs. Fin. Elem. Anal. Des. 228, 104047 (2024). DOI https://doi.org/10.1016/j.finel.2023.104047

    work page doi:10.1016/j.finel.2023.104047 2024

  13. [13]

    Deparis, S.: Reduced basis error bound computation of parameter-dependent Navier– Stokes equations by the natural norm approach. SIAM J. Numer. Anal.46(4), 2039– 2067 (2008). DOI https://doi.org/10.1137/060674181

    work page doi:10.1137/060674181 2039

  14. [14]

    Nonlinearity35(1), 278–310 (2021)

    E, W., Han, J., Jentzen, A.: Algorithms for solving high dimensional PDEs: From nonlinear Monte Carlo to machine learning. Nonlinearity35(1), 278–310 (2021). DOI https://doi.org/10.1088/1361-6544/ac337f

    work page doi:10.1088/1361-6544/ac337f 2021

  15. [15]

    E, W., Yu, B.: The Deep Ritz method: A deep learning-based numerical algorithm for solving variational problems. Commun. Math. Stat.6(1), 1–12 (2018). DOI https: //doi.org/10.1007/s40304-018-0127-z

    work page doi:10.1007/s40304-018-0127-z 2018

  16. [16]

    Flyer, N., Wright, G.: Transport schemes on a sphere using radial basis functions. J. Comput. Phys.226(1), 1059–1084 (2007). DOI https://doi.org/10.1016/j.jcp.2007.05. 009

    work page doi:10.1016/j.jcp.2007.05 2007

  17. [17]

    AIAA J.9(1), 177–179 (1971)

    Fox, R., Miura, H.: An approximate analysis technique for design calculations. AIAA J.9(1), 177–179 (1971). DOI https://doi.org/10.2514/3.49225

    work page doi:10.2514/3.49225 1971

  18. [18]

    Fresca, S., Manzoni, A.: POD-DL-ROM: Enhancing deep learning-based reduced order models for nonlinear parametrized PDEs by proper orthogonal decomposition. Comput. Methods Appl. Mech. Eng.388, 114181 (2022). DOI https://doi.org/10.1016/j.cma. 2021.114181

    work page doi:10.1016/j.cma 2022

  19. [19]

    Part I—open-loop model development

    Graham, W., Peraire, J., Tang, K.: Optimal control of vortex shedding using low- order models. Part I—open-loop model development. Int. J. Numer. Methods Eng. 44(7), 945–972 (1999). DOI https://doi.org/10.1002/(SICI)1097-0207(19990310)44: 7⟨945::AID-NME537⟩3.0.CO;2-F Title Suppressed Due to Excessive Length 27

    work page doi:10.1002/(sici)1097-0207(19990310)44: 1999

  20. [20]

    ESAIM: M2AN41(3), 575–605 (2007)

    Grepl, M., Maday, Y., Nguyen, N., Patera, A.: Efficient reduced-basis treatment of nonaffine and nonlinear partial differential equations. ESAIM: M2AN41(3), 575–605 (2007). DOI https://doi.org/10.1051/m2an:2007031

    work page doi:10.1051/m2an:2007031 2007

  21. [21]

    In: Real-Time PDE-Constrained Optimization, pp

    Grepl, M., Nguyen, N., Veroy, K., Patera, A., Liu, G.: Certified Rapid Solution of Partial Differential Equations for Real-Time Parameter Estimation and Optimization. In: Real-Time PDE-Constrained Optimization, pp. 199–216. SIAM, Philadelphia (2007). DOI https://doi.org/10.1137/1.9780898718935.ch10

    work page doi:10.1137/1.9780898718935.ch10 2007

  22. [22]

    Gudmundsson, K., Colonius, T.: Instability wave models for the near-field fluctuations of turbulent jets. J. Fluid Mech.689, 97–128 (2011). DOI https://doi.org/10.1017/jfm. 2011.401

    work page doi:10.1017/jfm 2011

  23. [23]

    Elsevier, Amsterdam (2012)

    Gunzburger, M.: Finite Element Methods for Viscous Incompressible Flows: A Guide to Theory, Practice, and Algorithms. Elsevier, Amsterdam (2012)

    2012

  24. [24]

    Gunzburger, M., Peterson, J., Shadid, J.: Reduced-order modeling of time-dependent PDEs with multiple parameters in the boundary data. Comput. Methods Appl. Mech. Eng.196(4-6), 1030–1047 (2007). DOI https://doi.org/10.1016/j.cma.2006.08.004

    work page doi:10.1016/j.cma.2006.08.004 2007

  25. [25]

    Model Reduction and Approximation: Theory and Algorithms15, 65 (2017)

    Haasdonk, B.: Reduced basis methods for parametrized PDEs—a tutorial introduction for stationary and instationary problems. Model Reduction and Approximation: Theory and Algorithms15, 65 (2017)

    2017

  26. [26]

    Han, J., Jentzen, A., E, W.: Deep learning-based numerical methods for high- dimensional parabolic partial differential equations and backward stochastic differential equations. Commun. Math. Stat.5(4), 349–380 (2017). DOI https://doi.org/10.1007/ s40304-017-0117-6

    2017

  27. [27]

    Han, J., Jentzen, A., E, W.: Solving high-dimensional partial differential equations using deep learning. Proc. Natl. Acad. Sci. USA115(34), 8505–8510 (2018). DOI https: //doi.org/10.1073/pnas.1718942115

    work page doi:10.1073/pnas.1718942115 2018

  28. [28]

    Hannachi, A., Jolliffe, I., Stephenson, D.: Empirical orthogonal functions and related techniques in atmospheric science: A review. Int. J. Climatol.27(9), 1119–1152 (2007). DOI https://doi.org/10.1002/joc.1499

    work page doi:10.1002/joc.1499 2007

  29. [29]

    Heinrichs, W.: A stabilized treatment of the biharmonic operator with spectral methods. SIAM J. Sci. Stat. Comput.12(5), 1162–1172 (1991). DOI https://doi.org/10.1137/ 0912061

    1991

  30. [30]

    Springer, Cham (2016)

    Hesthaven, J., Rozza, G., Stamm, B.: Certified Reduced Basis Methods for Parametrized Partial Differential Equations. Springer, Cham (2016). DOI https://doi.org/10.1007/ 978-3-319-22470-1

    2016

  31. [31]

    Alicante, A

    Iapichino, L., Rozza, G., Quarteroni, A.: Reduced basis method and domain decom- position for elliptic problems in multi-parameter domains. In: Numerical Mathematics and Advanced Applications 2011, pp. 367–374 (2013). DOI https://doi.org/10.1016/j. camwa.2015.12.001

    work page doi:10.1016/j 2011

  32. [32]

    Ito, K., Ravindran, S.: A reduced-order method for simulation and control of fluid flows. J. Comput. Phys.143(2), 403–425 (1998). DOI https://doi.org/10.1006/jcph.1998.5943

    work page doi:10.1006/jcph.1998.5943 1998

  33. [33]

    Ito, K., Ravindran, S.: Reduced basis method for optimal control of unsteady viscous flows. Int. J. Comput. Fluid Dyn.15(2), 97–113 (2001). DOI https://doi.org/10.1080/ 10618560108970021

    2001

  34. [34]

    Ito, K., Schroeter, J.: Reduced order feedback synthesis for viscous incompressible flows. Math. Comput. Model.33(1-3), 173–192 (2001). DOI https://doi.org/10.1016/ S0895-7177(00)00237-5

    2001

  35. [35]

    arXiv preprint arXiv:2505.08214 (2025)

    Jin, T., Peng, Z., Xiang, Y.: Adaptive and hybrid reduced order models to miti- gate Kolmogorov barrier in a multiscale kinetic transport equation. arXiv preprint arXiv:2505.08214 (2025). DOI https://doi.org/10.48550/arXiv.2505.08214

    work page doi:10.48550/arxiv.2505.08214 2025

  36. [36]

    NSPOD: Accelerating Krylov solvers via DeepONet-learned POD subspaces

    Levrero-Florencio, F., Lee, Y., Pathak, J., Karniadakis, G.: NSPOD: Neural Subspace Proper Orthogonal Decomposition for dynamical systems on complex geometries. arXiv preprint arXiv:2605.07828 (2026). DOI https://doi.org/10.48550/arXiv.2605.07828

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2605.07828 2026

  37. [37]

    arXiv preprint arXiv:2511.07684 (2025)

    Li, J., Bespalov, A., Li, J.: Nemytskii neural operator: A nonlinear model reduction method for parametrized partial differential equations. arXiv preprint arXiv:2511.07684 (2025). DOI https://doi.org/10.48550/arXiv.2511.07684

    work page doi:10.48550/arxiv.2511.07684 2025

  38. [38]

    Li, X., Fan, Y., Wang, C., Yu, X., Sun, S., Sun, D.: A POD-DEIM reduced model for compressible gas reservoir flow based on the Peng–Robinson equation of state. J. Nat. Gas Sci. Eng.79, 103367 (2020). DOI https://doi.org/10.1016/j.jngse.2020.103367 28 Tianhao Hu, Zecheng Gan

    work page doi:10.1016/j.jngse.2020.103367 2020

  39. [39]

    Fourier Neural Operator for Parametric Partial Differential Equations

    Li, Z., Kovachki, N., Azizzadenesheli, K., Liu, B., Bhattacharya, K., Stuart, A., Anand- kumar, A.: Fourier neural operator for parametric partial differential equations. arXiv preprint arXiv:2010.08895 (2020). DOI https://doi.org/10.48550/arXiv.2010.08895

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2010.08895 2010

  40. [40]

    Springer, New York (1977)

    Lo` eve, M.: Probability Theory I. Springer, New York (1977). DOI https://doi.org/10. 1007/978-1-4684-9464-8

    1977

  41. [41]

    Lu, L., Jin, P., Pang, G., Zhang, Z., Karniadakis, G.: Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators. Nat. Mach. Intell.3(3), 218–229 (2021). DOI https://doi.org/10.1038/s42256-021-00302-5

    work page doi:10.1038/s42256-021-00302-5 2021

  42. [42]

    Lumley, J.: The structure of inhomogeneous turbulent flows. Atmos. Turbul. Radio Propag. pp. 166–178 (1967)

    1967

  43. [43]

    Elsevier, New York (2012)

    Lumley, J.: Stochastic Tools in Turbulence. Elsevier, New York (2012)

    2012

  44. [44]

    arXiv preprint arXiv:2208.12630 (2022)

    Mendez, M.: Generalized and multiscale modal analysis. arXiv preprint arXiv:2208.12630 (2022). DOI https://doi.org/10.48550/arXiv.2208.12630

    work page doi:10.48550/arxiv.2208.12630 2022

  45. [45]

    Mendez, M., Balabane, M., Buchlin, J.M.: Multi-scale proper orthogonal decomposition of complex fluid flows. J. Fluid Mech.870, 988–1036 (2019). DOI https://doi.org/10. 1017/jfm.2019.212

    2019

  46. [46]

    Mendez, M., Hess, D., Watz, B., Buchlin, J.M.: Multiscale proper orthogonal decom- position (mPOD) of TR-PIV data—a case study on stationary and transient cylinder wake flows. Meas. Sci. Technol.31(9), 094014 (2020). DOI https://doi.org/10.1088/ 1361-6501/ab82be

    2020

  47. [47]

    Mendez, M., Scelzo, M., Buchlin, J.M.: Multiscale modal analysis of an oscillating im- pinging gas jet. Exp. Therm. Fluid Sci.91, 256–276 (2018). DOI https://doi.org/10. 1016/j.expthermflusci.2017.10.032

    2018

  48. [48]

    Oliveira, I., Patera, A.: Reduced-basis techniques for rapid reliable optimization of sys- tems described by affinely parametrized coercive elliptic partial differential equations. Optim. Eng.8(1), 43–65 (2007). DOI https://doi.org/10.1007/s11081-007-9002-6

    work page doi:10.1007/s11081-007-9002-6 2007

  49. [49]

    Patera, A.: A spectral element method for fluid dynamics: Laminar flow in a channel expansion. J. Comput. Phys.54(3), 468–488 (1984). DOI https://doi.org/10.1016/ 0021-9991(84)90128-1

    1984

  50. [50]

    MIT, Cambridge (2007)

    Patera, A., Rozza, G.: Reduced Basis Approximation and A Posteriori Error Estimation for Parametrized Partial Differential Equations. MIT, Cambridge (2007). Lecture Notes

    2007

  51. [51]

    Springer Science & Business Media, New York (2012)

    Pinkus, A.: n-Widths in Approximation Theory. Springer Science & Business Media, New York (2012)

    2012

  52. [52]

    Politis, A.: Microphone array processing for parametric spatial audio techniques. Ph.D. thesis, Aalto University (2016)

    2016

  53. [53]

    Prud’homme, C., Rovas, D., Veroy, K., Machiels, L., Maday, Y., Patera, A., Turinici, G.: Reliable real-time solution of parametrized partial differential equations: Reduced-basis output bound methods. J. Fluids Eng.124(1), 70–80 (2002). DOI https://doi.org/10. 1115/1.1448332

    2002

  54. [54]

    Quarteroni, A., Manzoni, A., Negri, F.: Reduced Basis Methods for Partial Differential Equations: An Introduction, vol. 92. Springer, Cham (2015). DOI doi.org/10.1007/ 978-3-319-15431-2

    2015

  55. [55]

    Raissi, M., Perdikaris, P., Karniadakis, G.: Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J. Comput. Phys.378, 686–707 (2019). DOI https://doi.org/10. 1016/j.jcp.2018.10.045

    2019

  56. [56]

    Rozza, G., Huynh, D., Patera, A.: Reduced basis approximation and a posteriori error estimation for affinely parametrized elliptic coercive partial differential equations: Ap- plication to transport and continuum mechanics. Arch. Comput. Methods Eng.15(3), 229–275 (2008). DOI https://doi.org/10.1007/s11831-008-9019-9

    work page doi:10.1007/s11831-008-9019-9 2008

  57. [57]

    San, O., Borggaard, J.: Principal interval decomposition framework for POD reduced- order modeling of convective Boussinesq flows. Int. J. Numer. Methods Fluids78(1), 37–62 (2015). DOI https://doi.org/10.1002/fld.4006

    work page doi:10.1002/fld.4006 2015

  58. [58]

    Schmidt, O., Towne, A., Colonius, T., Cavalieri, A., Jordan, P., Br` es, G.: Wavepackets and trapped acoustic modes in a turbulent jet: Coherent structure eduction and global stability. J. Fluid Mech.825, 1153–1181 (2017). DOI https://doi.org/10.1017/jfm.2017. 407 Title Suppressed Due to Excessive Length 29

    work page doi:10.1017/jfm.2017 2017

  59. [59]

    Direct solvers of second- and fourth- order equations using Chebyshev polynomials

    Shen, J.: Efficient spectral-Galerkin method II. Direct solvers of second- and fourth- order equations using Chebyshev polynomials. SIAM J. Sci. Comput.16(1), 74–87 (1995). DOI https://doi.org/10.1137/0916006

    work page doi:10.1137/0916006 1995

  60. [60]

    Sieber, M., Paschereit, C., Oberleithner, K.: Spectral proper orthogonal decomposition. J. Fluid Mech.792, 798–828 (2016). DOI https://doi.org/10.1017/jfm.2016.103

    work page doi:10.1017/jfm.2016.103 2016

  61. [61]

    Sinha, A., Rodr´ ıguez, D., Br` es, G., Colonius, T.: Wavepacket models for supersonic jet noise. J. Fluid Mech.742, 71–95 (2014). DOI https://doi.org/10.1017/jfm.2013.660

    work page doi:10.1017/jfm.2013.660 2014

  62. [62]

    Strang, G.: On the construction and comparison of difference schemes. SIAM J. Numer. Anal.5(3), 506–517 (1968). DOI https://doi.org/10.1137/0705041

    work page doi:10.1137/0705041 1968

  63. [63]

    AIAA J.55(12), 4013–4041 (2017)

    Taira, K., Brunton, S., Dawson, S., Rowley, C., Colonius, T., McKeon, B., Schmidt, O., Gordeyev, S., Theofilis, V., Ukeiley, L.: Modal analysis of fluid flows: An overview. AIAA J.55(12), 4013–4041 (2017). DOI https://doi.org/10.2514/1.J056060

    work page doi:10.2514/1.j056060 2017

  64. [64]

    Towne, A., Schmidt, O., Colonius, T.: Spectral proper orthogonal decomposition and its relationship to dynamic mode decomposition and resolvent analysis. J. Fluid Mech. 847, 821–867 (2018). DOI https://doi.org/10.1017/jfm.2018.283

    work page doi:10.1017/jfm.2018.283 2018

  65. [65]

    SIAM, Philadelphia (2000)

    Trefethen, L.: Spectral Methods in MATLAB. SIAM, Philadelphia (2000). DOI https: //doi.org/10.1137/1.9780898719598

    work page doi:10.1137/1.9780898719598 2000

  66. [66]

    Veroy, K., Patera, A.: Certified real-time solution of the parametrized steady incom- pressible Navier–Stokes equations: Rigorous reduced-basis a posteriori error bounds. Int. J. Numer. Meth. Fluids47(8-9), 773–788 (2005). DOI https://doi.org/10.1002/fld.867

    work page doi:10.1002/fld.867 2005

  67. [67]

    In: 16th AIAA Computational Fluid Dynamics Conference, p

    Veroy, K., Prud’homme, C., Rovas, D., Patera, A.: A posteriori error bounds for reduced- basis approximation of parametrized noncoercive and nonlinear elliptic partial differ- ential equations. In: 16th AIAA Computational Fluid Dynamics Conference, p. 3847 (2003). DOI https://doi.org/10.2514/6.2003-3847

    work page doi:10.2514/6.2003-3847 2003

  68. [68]

    Lecture Notes, University of Konstanz4(4), 1–29 (2013)

    Volkwein, S.: Proper orthogonal decomposition: Theory and reduced-order modelling. Lecture Notes, University of Konstanz4(4), 1–29 (2013)

    2013

  69. [69]

    Wang, B., Xu, Y., Zhou, Y., Xue, D.: Multi-fidelity learning in materials informatics: Methodologies, applications, and outlook. J. Mater. Inform.6, 16 (2026). DOI https: //doi.org/10.20517/jmi.2025.85

    work page doi:10.20517/jmi.2025.85 2026

  70. [70]

    Open Phys.14(1), 588–601 (2016)

    Wang, Q., Yu, X., Sun, S.: POD-Galerkin model for incompressible single-phase flow in porous media. Open Phys.14(1), 588–601 (2016). DOI https://doi.org/10.1515/ phys-2016-0061

    2016

  71. [71]

    Wang, Z., Xiao, D., Fang, F., Govindan, R., Pain, C., Guo, Y.: Model identification of reduced order fluid dynamics systems using deep learning. Int. J. Numer. Methods Fluids86(4), 255–268 (2018). DOI https://doi.org/10.1002/fld.4416

    work page doi:10.1002/fld.4416 2018

  72. [72]

    Xiao, D., Fang, F., Buchan, A., Pain, C., Navon, I., Muggeridge, A.: Non-intrusive reduced order modelling of the Navier–Stokes equations. Comput. Methods Appl. Mech. Eng.293, 522–541 (2015). DOI https://doi.org/10.1016/j.cma.2015.05.015

    work page doi:10.1016/j.cma.2015.05.015 2015