Pith. sign in

REVIEW 4 major objections 5 minor 33 references

A parallel implementation of reduced-order modeling of large-scale systems

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A fully parallelized Operator Inference procedure builds physics-based reduced-order models from snapshot data with state dimensions too large for a single computer, demonstrated on a 2D cylinder-flow benchmark where a 10-mode ROM trained…

desk verdict A useful, well-written tutorial for the existing dOpInf algorithm, with correct math and working code, but the scalability claims in the abstract outrun the evidence presented. read the letter →

arxiv 2504.14338 v1 pith:CCJXVUOH submitted 2025-04-19 cs.DC cs.CE

Ionut-Gabriel Farcas , Rayomand P. Gundevia , Ramakanth Munipalli , Karen E. Willcox This is my paper · ORCID
classification cs.DCcs.CE MSC 65Y0565F1565F2068W10
keywords distributedOperatorInferencereduced-ordermodelingproperorthogonaldecompositionmethodofsnapshotsparallelalgorithmstall-and-skinnydataregularizedleastsquaresMPI
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper describes Distributed Operator Inference (dOpInf), a complete parallel pipeline for turning high-dimensional simulation data into cheap physics-based reduced-order models. Its central claim is that every stage—loading data, transformations, dimensionality reduction, operator learning, and postprocessing—can be distributed so that datasets with millions of degrees of freedom, too large for one computer, can be processed on many cores. The algorithm avoids ever constructing the full reduced basis: it builds an $n_t \times n_t$ Gram matrix, diagonalizes it, and forms the projected snapshots from that small matrix alone. In the paper's benchmark, a 10-dimensional quadratic ROM trained on 600 snapshots over $[4,7]$ seconds predicts the 2D cylinder wake over $[7,10]$ seconds, with measured speedups up to $p=8$ cores. If the claims hold, dOpInf makes tasks like design exploration, risk assessment, and uncertainty quantification feasible for simulations that currently cannot be reduced on a workstation.

What carries the argument

The load-bearing object is the Gram-matrix eigen-decomposition identity $\mathbf{D}=\mathbf{Q}^\top\mathbf{Q}=\mathbf{W}\boldsymbol{\Sigma}^2\mathbf{W}^\top$, combined with $\hat{\mathbf{Q}}=\mathbf{T}_r^\top\mathbf{D}$ where $\mathbf{T}_r=\mathbf{U}_r\boldsymbol{\Lambda}_r^{-1/2}$. It lets the algorithm compute the low-dimensional projected snapshots and the local POD-basis components $\mathbf{V}_{r,i}=\mathbf{Q}_i\mathbf{T}_r$ using only dense matrix-matrix products and a symmetric eigendecomposition of size $n_t\times n_t$, never the $n\times n_t$ SVD. All data-dependent operations act on row blocks (loading, centering, scaling, local Gram matrices, local basis components), so communication is confined to one Allreduce of the Gram matrix and one Allreduce for the optimal regularization pair.

What would settle it

Construct a dataset with $n_t$ large enough that each rank's copy of $\mathbf{D}$ approaches its memory limit (say $n_t=10^5$, requiring an 80 GB dense matrix in double precision), and run the dOpInf steps: the Allreduce of local Gram matrices and the redundant symmetric eigendecomposition will dominate the runtime, removing the speedup and disproving the claim that all elements are scalable.

Watch

Extended reading notes

Core claim

The central discovery is an identity that removes the large state dimension from the expensive part of data-driven model reduction. Starting from the snapshot matrix $\mathbf{S}\in\mathbb{R}^{n\times n_t}$, each rank forms a local Gram matrix $\mathbf{D}_i=\mathbf{Q}_i^\top\mathbf{Q}_i$; a single Allreduce sums them into $\mathbf{D}=\mathbf{Q}^\top\mathbf{Q}$. Because $\mathbf{D}=\mathbf{W}\boldsymbol{\Sigma}^2\mathbf{W}^\top$, the eigenvectors of $\mathbf{D}$ are the right singular vectors and its eigenvalues are the squared singular values, so the rank-$r$ POD basis need never be formed explicitly. Instead the projected data are $\hat{\mathbf{Q}}=\mathbf{T}_r^\top\mathbf{D}$ with $\mathbf{T}_r=\mathbf{U}_r\boldsymbol{\Lambda}_r^{-1/2}$, a product of two small $n_t\times r$ and $n_t\times n_t$ matrices. The paper then learns the reduced operators of a discrete quadratic ROM by solving a regularized least-squares problem in the reduced space, and reconstructs solutions at probe locations without ever lifting the full state. The claim is that this is a complete, scalable distributed procedure: in the example, 600 downsampled snapshots of a $292{,}678$-dimensional flow yield a 10-mode ROM whose forecast matches the reference at three probe locations.

Load-bearing premise

The whole procedure assumes tall-and-skinny data: the number of snapshots $n_t$ is small enough that every MPI rank can store, Allreduce, and eigendecompose the $n_t\times n_t$ Gram matrix, and that all data transformations act pointwise on row blocks without communication.

Editorial extensions

If this is right

  • Training data with state dimension $n\sim 10^6$–$10^9$ can be processed on distributed-memory machines, since no single rank ever stores the full snapshot matrix or forms the full POD basis.
  • Because the reduced model is quadratic and inferred from data, predictions beyond the training horizon are possible; in the demonstration a 10-mode ROM trained on $[4,7]$ s accurately tracks the cylinder wake through $[7,10]$ s.
  • Regularized least squares with a grid search over two hyperparameters keeps the inferred operators from overfitting, and the grid search parallelizes over ranks.
  • Measured strong scaling on the small benchmark is nearly linear up to $p=4$ and degrades at $p=8$; the paper reports near-ideal speedups up to $2{,}048$ cores in the larger rotating-detonation application [1].

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same Gram-matrix identity transfers to other data-driven reduced-modeling methods that the paper names as compatible—DMD, quadratic manifolds, parametric ROMs—so the parallel strategy is not tied to Operator Inference alone.
  • The tall-and-skinny assumption is the real boundary: if $n_t$ grows so the $n_t\times n_t$ Gram matrix cannot be stored or Allreduced on every rank, the redundant eigendecomposition and $O(n_t^2)$ communication will dominate; a testable extension is a distributed eigendecomposition or randomized low-rank approximation of $\mathbf{D}$.
  • Because the projected snapshots are formed from $\mathbf{D}$ alone, the algorithm could be adapted to streaming or incremental updates: new snapshots only update local Gram matrices, then one Allreduce.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper presents dOpInf, a distributed implementation of Operator Inference for constructing physics-based reduced-order models from large-scale snapshot data, in a tutorial format illustrated on two-dimensional transient flow past a cylinder. The algorithm is described in five steps: parallel training-data loading, parallel data transformations, parallel dimensionality reduction via the method of snapshots with a distributed Gram matrix, parallel reduced-operator learning through regularized least squares, and parallel postprocessing of the reduced solution. The numerical experiment trains a 10-dimensional ROM on the time interval [4,7] seconds and evaluates it on [7,10] seconds, reporting accurate probe predictions and strong-scaling measurements up to 8 cores. The paper claims that all elements of dOpInf are scalable and that the approach can scale to thousands of processors, with the large-scale scalability evidence deferred to a companion paper.

Significance. The tutorial is potentially valuable as a practical guide for the model-reduction community: it provides a complete, reproducible implementation with a publicly available repository, and the distributed method-of-snapshots derivation in Eqs. (5)-(8) is correct and elegant. The out-of-sample evaluation on a time horizon beyond the training interval is a genuine test of predictive performance. The strength of the paper lies in its didactic clarity and the availability of working code. However, the paper's headline scalability claims go beyond what the presented experiments and the algorithm design can support, and the hyperparameter-selection procedure partly uses the evaluation horizon, which tempers the force of the out-of-sample claim.

major comments (4)
  1. [Abstract, §I, §III.D, §IV] The abstract and §I state that 'all elements of dOpInf are scalable' and that the approach 'can scale to the thousands of processors available on leadership HPC platforms', but this is not established by the manuscript. In §III.D, Step III assembles the n_t-by-n_t Gram matrix via an Allreduce and then every rank performs a full dense eigendecomposition of that matrix (code lines 79-82); this requires O(n_t^2) memory per rank, O(n_t^2) communication, and O(n_t^3) redundant serial work. The manuscript correctly notes in the text that these computations are cheap only when n_t is small relative to the state dimension m, but the abstract and introduction state the scalability claim without this tall-and-skinny qualifier. The strong-scaling experiment stops at p=8 and shows deteriorating efficiency, with the leadership-scale claim deferred to Ref. [1]. I recommend qualifying the scalability claims explicitly to the tall-and-skinny regime and adding a complexity analysis that spells out the O(n_t^2) communication and O(n_t^3) per-rank eigendecomposition costs.
  2. [§III.E, §IV] The regularization selection in Step IV filters candidate (beta1, beta2) pairs using a bounded-growth criterion over a trial horizon that is stated to be the same as the full target horizon [4,10] seconds (code lines 197-202 and the associated text). Therefore, the ROM that is evaluated on [7,10] seconds was selected using information from that same evaluation interval. This does not make the training/evaluation split circular in the standard sense, because the operators are learned only from [4,7], but it does mean that the reported prediction accuracy is partly a consequence of hyperparameter filtering on the test horizon. I recommend selecting the regularization pair using only the training interval (or a validation subinterval), or explicitly reporting the sensitivity of the results to this choice, so that the out-of-sample claim in §IV is clean.
  3. [§III.F, code lines 323-325] The probe locations are inconsistent between the text and the code. The text in §III.F states that the three probe locations are (0.40, 0.20), (0.60, 0.20), and (1.00, 0.20) with grid-point indices {16,992; 48,250; 130,722}, while the code at lines 323-325 sets target_probe_indices = [48250, 77502, 130722]. Since the paper is a tutorial whose primary purpose is reproducibility, this mismatch is material and should be corrected in one place, or an explanation of the intended mapping should be provided.
  4. [§III.D, code line 90, §IV] The retained-energy threshold is stated as 99.95% in Eq. (9) and the surrounding text, but the code sets target_ret_energy = 0.9996 (line 90) and §IV reports '99.96%'. This inconsistency should be fixed so that the threshold quoted in the mathematical description, the code, and the results section agree.
minor comments (5)
  1. [§III.D, Eq. (4)] The text says the singular values are in 'non-decreasing order' while writing sigma_1 >= sigma_2 >= ...; this should be 'non-increasing order'.
  2. [§III.A, Remark 2] The phrase 'In dOpinf' uses inconsistent capitalization; it should be 'dOpInf' for consistency with the rest of the paper.
  3. [§IV, Figure 4] The speed-up discussion would be easier to interpret if the parallel efficiency (speedup divided by p) were reported alongside the CPU times, since the p=8 point corresponds to roughly 60% efficiency and the paper's phrase 'excellent up to p=4' is fine but should be quantified.
  4. [§III.E, code line 262] The implementation solves the OpInf least-squares problem via the normal equations with np.linalg.solve; the text mentions alternatives such as SVD or QR, but it would be helpful to add a sentence noting that the normal equations can be ill-conditioned when the data matrix is poorly conditioned, particularly because the quadratic terms can have widely varying magnitudes.
  5. [§IV, Figure 3] The caption of Figure 3 does not state which velocity component (u_x or u_y) is plotted at the three probe locations; the text should specify this, since the two components have different characteristic behavior.

Circularity Check

2 steps flagged · score 3.0 of 10

Core POD/OpInf derivation is self-contained, but the reported 'prediction' is mildly compromised by target-horizon hyperparameter selection and the thousand-core scalability claim rests on a self-citation.

  1. fitted input called prediction [Sec. III.E (Step IV, regularization selection around Eqs. (12) and code lines 274-294); prediction reported in Sec. IV]
    "The optimal hyperparameters are chosen to minimize the training error, subject to the constraint that the inferred reduced coefficient have bounded growth over a trial time horizon [tinit, ttrial] with ttrial >= tfinal [27, 28]. ... In our implementation, the trial time horizon is the same as the target horizon, that is, [4, 10] seconds."

    Each candidate (beta1, beta2) pair is scored by solving the discrete ROM over nt_p=1200 steps covering [4,10] and by the ratio max_diff_Qhat_trial / max_diff_Qhat_train, where max_diff_Qhat_trial is taken over the full simulated horizon (lines 274-294), i.e., including the [7,10] window that Sec. IV labels as 'predictions beyond training'. The selected hyperparameters therefore encode information about whether each candidate model stays bounded over the exact prediction window before that model's output is presented as an out-of-sample prediction. The pointwise probe trajectories are not fitted to [7,10] reference data, so this is a model-selection leak rather than a full fit, but the prediction is not a blind test.

  2. self citation load bearing [Abstract and Sec. IV (scalability discussion, Ref. [1])]
    "All elements of dOpInf are scalable, leading to a fully parallelized reduced modeling approach that can scale to the thousands of processors available on leadership high-performance computing platforms. ... Note that the deterioration in speed-up here is due to the relatively small size of the example; In Ref. [1], near-ideal speed-ups were observed for up to p=2,048 cores."

    The paper's own strong-scaling experiment stops at p=8 and shows deteriorating speedup (8.35s at p=1 to 1.72s at p=8); the 'thousands of processors' and 'near-ideal speed-ups up to p=2,048' statements are supported only by Ref. [1], a companion paper by the same four authors. The abstract's headline scalability claim is thus load-bearing on a self-citation rather than on evidence within this manuscript. This does not affect the POD/OpInf derivation or the benchmark predictions, which are independently evaluated.

full rationale

The core derivation chain is not circular. Step III uses the method of snapshots with D = Q^T Q = sum_i Q_i^T Q_i and the exact identities D = W Sigma^2 W^T and V_r = Q U_r Lambda_r^{-1/2}, yielding Qhat = T_r^T D (Eqs. 5-8); this is standard linear algebra, not an input-output tautology. Step IV infers the constant, linear, and quadratic operators from the projected training snapshots over [4,7] only, and Sec. IV evaluates the resulting r=10 ROM on [7,10] at three probe locations. No equation defines the target prediction in terms of the reported ROM output. Two qualifications keep this from a clean 0-2 score. First, the regularization selection in Step IV applies a bounded-growth filter over the full [4,10] target horizon, which includes the prediction window, so the reported prediction is not a fully blind out-of-sample evaluation. Second, the paper's own scaling data end at p=8 with deteriorating efficiency, and the 'thousands of processors' claim in the abstract is deferred to the self-cited companion Ref. [1]. The paper transparently acknowledges some limits ('These computations are cheap when nt << m and r << m'; the p=8 deterioration 'is due to the relatively small size of the example'), and the internal inconsistencies (0.9995 in Eq. (9) vs 0.9996 in the code; probe index sets {16,992;48,250;130,722} in Sec. III.F vs [48250,77502,130722] in the code) are correctness/reproducibility issues rather than circularity. Overall the central predictive modeling result remains independent, so a score of 3 reflects the mild target-horizon leak and the load-bearing self-citation without alleging that the derivation is equivalent to its inputs.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical entities, forces, or dimensions. The only adjustable quantities are hyperparameters of the ROM inference procedure (regularization weights, reduced dimension, retained-energy threshold, and growth tolerance), all fitted or chosen by hand on the training and trial horizons.

free parameters (5)
  • beta1 regularization weight = 7.19e-8
    Selected by grid search over 8 log-spaced values in [1e-10, 1] (code lines 198, 247-262) to minimize training error subject to bounded growth over the trial horizon.
  • beta2 regularization weight = 51.79
    Selected by grid search over 8 log-spaced values in [1e-4, 1e4] (code lines 199, 247-262) with the same criterion as beta1.
  • Reduced dimension r = 10
    Chosen as the smallest dimension satisfying the retained-energy threshold (Eq. (9), code line 95); directly sets the ROM size.
  • Retained energy threshold = 0.9996 in code, 0.9995 in text Eq. (9)
    User-specified target for POD mode selection; the paper states 99.95% in the introduction of Step III but the code and results section use 99.96%.
  • Growth tolerance max_growth = 1.2
    User-defined filter (code line 202) that discards candidate ROMs whose reduced coefficients grow more than 20% relative to training over the trial horizon.
assumptions (5)
  • domain assumption The snapshot matrix can be partitioned into row blocks without cross-rank dependencies for centering and POD computations.
    Used in Steps I-III (Sec. III.B-D) to compute local Gram matrices D_i exactly and sum them via Allreduce to obtain D = Q^T Q (Eq. 5).
  • domain assumption The number of snapshots n_t is small enough that every rank can store, Allreduce, and eigendecompose the n_t by n_t Gram matrix.
    Step III.D performs eigh(D_global) redundantly on all ranks; the paper targets tall-and-skinny data (n_t << m) but does not analyze the case of large n_t.
  • domain assumption Pressure is omitted because it is determined by the velocity field through incompressibility.
    Sec. II.B states this simplification, which underlies the quadratic velocity-only ROM (Eq. 10).
  • domain assumption The discrete quadratic ROM form (11) matches the structure of the governing equations after centering.
    Sec. III.E assumes a quadratic polynomial model with a constant term; this is justified for Navier-Stokes with no lifting needed.
  • standard math Standard linear algebra facts from the method of snapshots: eigenvectors of the Gram matrix equal right singular vectors, and the POD basis is recovered by Eq. (7).
    Used in Eqs. (5)-(8) as a substitute for computing the thin SVD directly.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A parallel implementation of reduced-order modeling of large-scale systems." pith.science (2026). https://pith.science/paper/CCJXVUOH

@misc{pith2026250414338,
  author       = {Pith},
  title        = {Pith review of: A parallel implementation of reduced-order modeling of large-scale systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CCJXVUOH}},
  note         = {Machine review of arXiv:2504.14338}
}
read the original abstract

Motivated by the large-scale nature of modern aerospace engineering simulations, this paper presents a detailed description of distributed Operator Inference (dOpInf), a recently developed parallel algorithm designed to efficiently construct physics-based reduced-order models (ROMs) for problems with large state dimensions. One such example is the simulation of rotating detonation rocket engines, where snapshot data generated by high-fidelity large-eddy simulations have many millions of degrees of freedom. dOpInf enables, via distributed computing, the efficient processing of datasets with state dimensions that are too large to process on a single computer, and the learning of structured physics-based ROMs that approximate the dynamical systems underlying those datasets. All elements of dOpInf are scalable, leading to a fully parallelized reduced modeling approach that can scale to the thousands of processors available on leadership high-performance computing platforms. The resulting ROMs are computationally cheap, making them ideal for key engineering tasks such as design space exploration, risk assessment, and uncertainty quantification. To illustrate the practical application of dOpInf, we provide a step-by-step tutorial using a 2D Navier-Stokes flow over a step scenario as a case study. This tutorial guides users through the implementation process, making dOpInf accessible for integration into complex aerospace engineering simulations.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 13 canonical work pages

  1. [1]

    Distributed computing for physics-based data-driven reduced modeling at scale: Application to a rotating detonation rocket engine,

    Farcas, I.-G., Gundevia, R. P., Munipalli, R., and Willcox, K. E., “Distributed computing for physics-based data-driven reduced modeling at scale: Application to a rotating detonation rocket engine,”Computer Physics Communications, Vol. 313, 2025, p. Distribution Statement A: Approved for Public Release; Distribution is Unlimited. PA# AFRL-2024-6483 17 10...

    work page arXiv 2025

  2. [2]

    Data-drivenoperatorinferencefornonintrusiveprojection-basedmodelreduction,

    Peherstorfer,B.,andWillcox,K.,“Data-drivenoperatorinferencefornonintrusiveprojection-basedmodelreduction,” Computer Methods in Applied Mechanics and Engineering, Vol. 306, 2016, pp. 196–215. https://doi.org/https://doi.org/10.1016/j.cma. 2016.03.025

    doi:10.1016/j.cma 2016

  3. [3]

    Learning Nonlinear Reduced Models from Data with Operator Inference,

    Kramer, B., Peherstorfer, B., and Willcox, K. E., “Learning Nonlinear Reduced Models from Data with Operator Inference,” Annual Review of Fluid Mechanics, Vol. 56, No. Volume 56, 2024, 2024, pp. 521–548. https://doi.org/https://doi.org/10.1146/ annurev-fluid-121021-025220

    work page 2024

  4. [4]

    Turbulenceandthedynamicsofcoherentstructuresparti: Coherentstructures,

    Sirovich,L.,“Turbulenceandthedynamicsofcoherentstructuresparti: Coherentstructures,” QuarterlyofAppliedMathematics , Vol. 45, No. 3, 1987, pp. 561–571

    work page 1987

  5. [5]

    A Domain Decomposition Approach to POD,

    Beattie, C. A., Borggaard, J., Gugercin, S., and Iliescu, T., “A Domain Decomposition Approach to POD,”Proceedings of the 45th IEEE Conference on Decision and Control, 2006, pp. 6750–6756. https://doi.org/10.1109/CDC.2006.377642

    work page arXiv 2006

  6. [6]

    Approximate partitioned method of snapshots for POD,

    Wang, Z., McBee, B., and Iliescu, T., “Approximate partitioned method of snapshots for POD,”Journal of Computational and Applied Mathematics, Vol. 307, 2016, pp. 374–384. https://doi.org/https://doi.org/10.1016/j.cam.2015.11.023

    doi:10.1016/j.cam.2015.11.023 2016

  7. [7]

    High-Performance SVD Partial Spectrum Computation,

    Keyes, D., Ltaief, H., Nakatsukasa, Y., and Sukkari, D., “High-Performance SVD Partial Spectrum Computation,”Proceedings oftheInternationalConferenceforHighPerformanceComputing,Networking,StorageandAnalysis ,AssociationforComputing Machinery, New York, NY, USA, 2023. https://doi.org/10.1145/3581784.3607109

    arXiv 2023

  8. [8]

    Communication-optimal Parallel and Sequential QR and LU Factorizations,

    Demmel, J., Grigori, L., Hoemmen, M., and Langou, J., “Communication-optimal Parallel and Sequential QR and LU Factorizations,” SIAM Journal on Scientific Computing, Vol. 34, No. 1, 2012, pp. A206–A239. https://doi.org/10.1137/ 080731992

    work page 2012

Show all 33 references
  1. [9]

    DirectQRfactorizationsfortall-and-skinnymatricesinMapReducearchitectures,

    Benson,A.R.,Gleich,D.F.,andDemmel,J.,“DirectQRfactorizationsfortall-and-skinnymatricesinMapReducearchitectures,” 2013 IEEE International Conference on Big Data, 2013, pp. 264–272. https://doi.org/10.1109/BigData.2013.6691583

    2013

  2. [10]

    Parallel data-driven decomposition algorithm for large-scale datasets: with application to transitional boundary layers,

    Sayadi, T., and Schmid, P. J., “Parallel data-driven decomposition algorithm for large-scale datasets: with application to transitional boundary layers,”Theoretical and Computational Fluid Dynamics, Vol. 30, No. 5, 2016, pp. 415–428. https://doi.org/10.1007/s00162-016-0385-x

    2016 doi

  3. [11]

    Dynamic mode decomposition of numerical and experimental data,

    Schmid, P. J., “Dynamic mode decomposition of numerical and experimental data,”Journal of Fluid Mechanics, Vol. 656, 2010, p. 5–28. https://doi.org/10.1017/S0022112010001217

    2010 doi

  4. [12]

    N., Brunton, S

    Kutz, J. N., Brunton, S. L., Brunton, B. W., and Proctor, J. L.,Dynamic Mode Decomposition, Society for Industrial and Applied Mathematics, Philadelphia, PA, 2016. https://doi.org/10.1137/1.9781611974508

    2016 doi

  5. [13]

    On dynamic mode decomposition: Theory and applications,

    Tu, J. H., Rowley, C. W., Luchtenburg, D. M., Brunton, S. L., and Kutz, J. N., “On dynamic mode decomposition: Theory and applications,”Journal of Computational Dynamics, Vol. 1, No. 2, 2014, pp. 391–421

    2014

  6. [14]

    An overview of Trilinos,

    Long, K. R., Tuminaro, R. S., Bartlett, R. A., Hoekstra, R. J., Phipps, E. T., Kolda, T. G., Lehoucq, R. B., Thornquist, H. K., Hu, J. J., Williams, A. B., Salinger, A. G., Howle, V. E., Pawlowski, R. P., Willenbring, J. M., and Heroux, M. A., “An overview of Trilinos,” 2003. ...

    2003 doi

  7. [15]

    Sequential Karhunen-Loeve basis extraction and its application to images,

    Levy, A., and Lindenbaum, M., “Sequential Karhunen-Loeve basis extraction and its application to images,”Proceedings 1998 International Conference on Image Processing. ICIP98 (Cat. No.98CB36269), Vol. 2, 1998, pp. 456–460 vol.2. https://doi.org/10.1109/ICIP.1998.723422

    1998

  8. [16]

    Online learning of quadratic manifolds from streaming data for nonlinear dimensionality reduction and nonlinear model reduction,

    Schwerdtner, P., Mohan, P., Pachalieva, A., Bessac, J., O’Malley, D., and Peherstorfer, B., “Online learning of quadratic manifolds from streaming data for nonlinear dimensionality reduction and nonlinear model reduction,” 2024. URL https: //arxiv.org/abs/2409.02703

    2024 arXiv

  9. [17]

    J., Copeland, D

    Choi, Y., Arrighi, W. J., Copeland, D. M., Anderson, R. W., and Oxberry, G. M., “libROM,” [Computer Software] https://doi.org/10.11578/dc.20190408.3, oct 2019. https://doi.org/10.11578/dc.20190408.3

    2019 doi

  10. [18]

    Pressio: Enabling projection-based model reduction for large-scale nonlinear dynamical systems,

    Rizzi, F., Blonigan, P. J., Parish, E. J., and Carlberg, K. T., “Pressio: Enabling projection-based model reduction for large-scale nonlinear dynamical systems,” , 2021. URL https://arxiv.org/abs/2003.07798

    2021 arXiv

  11. [19]

    pyMOR – Generic Algorithms and Interfaces for Model Order Reduction,

    Milk, R., Rave, S., and Schindler, F., “pyMOR – Generic Algorithms and Interfaces for Model Order Reduction,”SIAM Journal on Scientific Computing, Vol. 38, No. 5, 2016, pp. S194–S216. https://doi.org/10.1137/15M1026614. Distribution Statement A: Approved for Public Release; Di...

    2016 doi

  12. [20]

    Unlocking massively parallel spectral proper orthogonal decompositions in the PySPOD package,

    Rogowski, M., Yeung, B. C., Schmidt, O. T., Maulik, R., Dalcin, L., Parsani, M., and Mengaldo, G., “Unlocking massively parallel spectral proper orthogonal decompositions in the PySPOD package,”Computer Physics Communications, 2024, p. 109246. https://doi.org/https://doi.org/1...

    2024

  13. [21]

    Operator inference for non-intrusive model reduction with quadratic manifolds,

    Geelen, R., Wright, S., and Willcox, K., “Operator inference for non-intrusive model reduction with quadratic manifolds,” Computer Methods in Applied Mechanics and Engineering, Vol. 403, 2023, p. 115717. https://doi.org/10.1016/j.cma.2022. 115717

    2023 doi

  14. [22]

    Quadratic approximation manifold for mitigating the Kolmogorov barrier in nonlinear projection- based model order reduction,

    Barnett, J., and Farhat, C., “Quadratic approximation manifold for mitigating the Kolmogorov barrier in nonlinear projection- based model order reduction,”Journal of Computational Physics, Vol. 464, 2022, p. 111348

    2022

  15. [23]

    A Survey of Projection-Based Model Reduction Methods for Parametric Dynamical Systems,

    Benner, P., Gugercin, S., and Willcox, K., “A Survey of Projection-Based Model Reduction Methods for Parametric Dynamical Systems,”SIAM Review, Vol. 57, No. 4, 2015, pp. 483–531. https://doi.org/10.1137/130932715

    2015 doi

  16. [24]

    Large-scale distributed linear algebra with tensor processing units,

    Lewis, A. G. M., Beall, J., Ganahl, M., Hauru, M., Mallick, S. B., and Vidal, G., “Large-scale distributed linear algebra with tensor processing units,”Proceedings of the National Academy of Sciences, Vol. 119, No. 33, 2022, p. e2122762119. https://doi.org/10.1073/pnas.2122762119

    2022 doi

  17. [25]

    Lift & Learn: Physics-informed machine learning for large- scale nonlinear dynamical systems,

    Qian, E., Kramer, B., Peherstorfer, B., and Willcox, K., “Lift & Learn: Physics-informed machine learning for large- scale nonlinear dynamical systems,”Physica D: Nonlinear Phenomena, Vol. 406, 2020, p. 132401. https://doi.org/https: //doi.org/10.1016/j.physd.2020.132401

    2020

  18. [26]

    Learning Physics-Based Reduced-Order Models for a Single-Injector Combustion Process,

    Swischuk, R., Kramer, B., Huang, C., and Willcox, K., “Learning Physics-Based Reduced-Order Models for a Single-Injector Combustion Process,”AIAA Journal, Vol. 58, No. 6, 2020, pp. 2658–2672. https://doi.org/10.2514/1.J058943

    2020 doi

  19. [27]

    Data-driven reduced-order models via regularised Operator Inference for a single-injector combustion process,

    McQuarrie, S. A., Huang, C., and Willcox, K. E., “Data-driven reduced-order models via regularised Operator Inference for a single-injector combustion process,”Journal of the Royal Society of New Zealand, Vol. 51, No. 2, 2021, pp. 194–211. https://doi.org/10.1080/03036758.2020.1863237

    2021

  20. [28]

    Reduced Operator Inference for Nonlinear Partial Differential Equations,

    Qian, E., Farcaş, I.-G., and Willcox, K., “Reduced Operator Inference for Nonlinear Partial Differential Equations,”SIAM Journal on Scientific Computing, Vol. 44, No. 4, 2022, pp. A1934–A1959. https://doi.org/10.1137/21M1393972

    2022 doi

  21. [29]

    H., and Van Loan, C

    Golub, G. H., and Van Loan, C. F.,Matrix Computations, 3rd ed., The Johns Hopkins University Press, 1996

    1996

  22. [30]

    Finding Structure with Randomness: Probabilistic Algorithms for Constructing Approximate Matrix Decompositions,

    Halko, N., Martinsson, P. G., and Tropp, J. A., “Finding Structure with Randomness: Probabilistic Algorithms for Constructing Approximate Matrix Decompositions,”SIAM Review, Vol. 53, No. 2, 2011, pp. 217–288. https://doi.org/10.1137/090771806

    2011 doi

  23. [31]

    Fast low-rank modifications of the thin singular value decomposition,

    Brand, M., “Fast low-rank modifications of the thin singular value decomposition,”Linear Algebra and its Applications, Vol. 415, No. 1, 2006, pp. 20–30. https://doi.org/https://doi.org/10.1016/j.laa.2005.07.021

    2006 doi

  24. [32]

    The Proper Orthogonal Decomposition in the Analysis of Turbulent Flows,

    Berkooz, G., Holmes, P., and Lumley, J. L., “The Proper Orthogonal Decomposition in the Analysis of Turbulent Flows,” Annual Review of Fluid Mechanics, Vol. 25, No. 1, 1993, pp. 539–575. https://doi.org/10.1146/annurev.fl.25.010193.002543

    1993

  25. [33]

    On filtering in non-intrusive data-driven reduced-order modeling,

    Farcas, I., Munipalli, R., and Willcox, K. E., “On filtering in non-intrusive data-driven reduced-order modeling,”AIAA AVIATION 2022 Forum, 2022. https://doi.org/10.2514/6.2022-3487. Distribution Statement A: Approved for Public Release; Distribution is Unlimited. PA# AFRL-202...

    2022 doi

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.