pith. sign in

arxiv: 2605.26025 · v1 · pith:GHCVDKJOnew · submitted 2026-05-25 · 🧮 math.NA · cs.NA

A Dynamic Subspace Approach for Low-rank Approximation of Large-scale Nonlinear Systems

Jack DeChant , Rudy Geelen , Shane A. McQuarrie , Johann Guilleminot This is my paper

Pith reviewed 2026-06-29 20:19 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords dynamic subspaceGrassmannian manifoldgeodesic pathlow-rank approximationtransport-dominated phenomenanonlinear systemsreduced-order modeling
0
0 comments X

The pith

Dynamic subspaces parameterized as geodesics on the Grassmannian achieve higher accuracy than static bases for low-rank approximation of transport-dominated systems at the same rank.

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

The paper presents a method that represents the evolving low-dimensional basis for a nonlinear system as a continuous geodesic trajectory on the Grassmannian manifold. This parameterization separates the geometric motion of the subspace from the underlying state dynamics, so the basis can follow changing features without the rank growth that static linear reductions require. Optimization occurs inside a reduced feature space whose size does not depend on the original high-dimensional state, keeping the cost manageable for large problems. The approach is demonstrated on a one-dimensional transport equation and a turbulent airfoil wake, where it maintains accuracy at lower ranks than fixed-subspace methods.

Core claim

The central claim is that a low-dimensional basis can be parameterized as a geodesic path on the Grassmannian manifold so that its time evolution decouples from the intrinsic state evolution. This dynamic tracking avoids the rank inflation needed by static low-rank approximations to capture transport-dominated or non-stationary physics. Performing the associated optimization in a reduced feature space renders computational cost independent of the full state dimension. Numerical tests on a 1D transport equation and a large-scale turbulent airfoil wake confirm higher accuracy than static linear approximations at identical ranks.

What carries the argument

Parameterization of the low-dimensional basis as a geodesic path on the Grassmannian manifold, which decouples subspace drift from state evolution and enables adaptive tracking without rank growth.

If this is right

  • Dynamic subspaces maintain accuracy for transport-dominated flows without the rank growth required by static bases.
  • Computational expense stays independent of full state dimension through reduced-feature-space optimization.
  • The method applies directly to large-scale nonlinear systems such as fluid wakes.
  • It supplies a scalable route to low-rank modeling of non-stationary dynamics.

Where Pith is reading between the lines

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

  • The same geodesic construction could be tested on other manifold-valued data reductions where features move across the domain.
  • If the decoupling holds, it may reduce the need for online basis adaptation techniques in existing reduced-order models.
  • Extension to three-dimensional flows would test whether the feature-space reduction continues to control cost at higher state dimensions.

Load-bearing premise

That geodesic parameterization on the Grassmannian lets the basis adapt to evolving physics without rank inflation and that optimization in reduced feature space keeps cost independent of original state dimension.

What would settle it

A direct comparison on the same transport problem showing that the dynamic method requires a rank increase comparable to the static method to reach a target accuracy, or that its error exceeds the static error at equal rank.

read the original abstract

We present a dynamic subspace approach for efficiently approximating large-scale systems by learning time-continuous trajectories on the Grassmannian manifold. By parameterizing a low-dimensional basis as a geodesic path, the method allows for adaptive tracking of evolving physics. Our approach decouples the geometric drift of the subspace from the intrinsic state evolution. This avoids the typical rank inflation required by static low-dimensional approximation methods to maintain accuracy, effectively breaking the Kolmogorov barrier in transport-dominated phenomena. To ensure scalability for high-dimensional data, the optimization is performed in a reduced feature space, rendering the computational cost independent of the large original state dimension. Numerical results for a 1D transport equation and a large-scale turbulent airfoil wake demonstrate that this dynamic subspace approach achieves higher accuracy than static linear approximations at equivalent ranks, positioning it as a robust and scalable method for the low-rank modeling of complex, non-stationary dynamical systems.

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

1 major / 0 minor

Summary. The paper presents a dynamic subspace approach for low-rank approximation of large-scale nonlinear systems by learning time-continuous trajectories on the Grassmannian manifold. The low-dimensional basis is parameterized as a geodesic path to adaptively track evolving physics, decoupling subspace drift from state evolution to avoid rank inflation in transport-dominated phenomena. Optimization occurs in a reduced feature space to render computational cost independent of the original state dimension. Numerical results on a 1D transport equation and turbulent airfoil wake are presented as demonstrating higher accuracy than static linear approximations at equivalent ranks.

Significance. If the accuracy claims hold with rigorous quantitative support, the work could advance reduced-order modeling for non-stationary systems by addressing the Kolmogorov barrier in transport problems via dynamic Grassmannian trajectories and scalable optimization.

major comments (1)
  1. Abstract: the claim that the dynamic subspace approach achieves higher accuracy than static linear approximations at equivalent ranks on the 1D transport equation and airfoil wake supplies no error metrics, implementation details, baselines, or derivations; the central claim cannot be evaluated from the provided text.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their review and the opportunity to clarify the manuscript. We respond to the single major comment below.

read point-by-point responses

  1. Referee: [—] Abstract: the claim that the dynamic subspace approach achieves higher accuracy than static linear approximations at equivalent ranks on the 1D transport equation and airfoil wake supplies no error metrics, implementation details, baselines, or derivations; the central claim cannot be evaluated from the provided text.

    Authors: We agree the abstract is a concise summary and does not embed specific quantitative metrics, implementation details, or derivations. The full manuscript supplies these in Sections 4.1 (1D transport equation) and 4.2 (turbulent airfoil wake), where relative L2-error tables compare the dynamic geodesic subspace against static POD baselines at matched ranks, together with the reduced-feature-space optimization procedure and geodesic derivations from Sections 2–3. The central accuracy claim is therefore supported by the body of the paper rather than the abstract alone. We can revise the abstract to include one or two representative quantitative statements if the editor requests. revision: partial

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper introduces a dynamic subspace method via geodesic parameterization on the Grassmannian, decoupling subspace drift from state evolution and performing optimization in a reduced feature space. The abstract and claims present this as a novel construction supported by direct numerical demonstrations on the 1D transport equation and airfoil wake, showing accuracy gains at fixed rank. No equations, derivations, or self-citations appear in the provided text that reduce any central claim to a fitted input, self-definition, or prior author result by construction. The approach is self-contained against external benchmarks with independent validation, yielding no load-bearing circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No free parameters, axioms, or invented entities are specified in the abstract.

pith-pipeline@v0.9.1-grok · 5691 in / 964 out tokens · 24616 ms · 2026-06-29T20:19:45.370008+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

39 extracted references · 36 canonical work pages

  1. [1]

    A survey of projection-based model reduction methods for parametric dynamical systems,

    Benner, P., Güğercin, 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

    work page doi:10.1137/130932715 2015

  2. [2]

    Turbulence and the dynamics of coherent structures. I. Coherent structures,

    Sirovich, L., “Turbulence and the dynamics of coherent structures. I. Coherent structures,”Quarterly of Applied Mathematics, Vol. 45, No. 3, 1987, pp. 561–571. https://doi.org/10.1090/qam/910462

    work page doi:10.1090/qam/910462 1987

  3. [3]

    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

    work page doi:10.1146/annurev.fl.25.010193.002543 1993

  4. [4]

    Model reduction of dynamical systems on nonlinear manifolds using deep convolutional autoencoders,

    Lee, K., and Carlberg, K. T., “Model reduction of dynamical systems on nonlinear manifolds using deep convolutional autoencoders,”Journal of Computational Physics, Vol. 404, 2020, p. 108973. https://doi.org/10.1016/j.jcp.2019.108973

    work page doi:10.1016/j.jcp.2019.108973 2020

  5. [5]

    A survey of model reduction by balanced truncation and some new results,

    Güğercin, S., and Antoulas, A. C., “A survey of model reduction by balanced truncation and some new results,”International Journal of Control, Vol. 77, No. 8, 2004, pp. 748–766. https://doi.org/10.1080/00207170410001713448

    work page doi:10.1080/00207170410001713448 2004

  6. [6]

    Data-driven balancing of linear dynamical systems,

    Gosea, I. V., Güğercin, S., and Beattie, C., “Data-driven balancing of linear dynamical systems,”SIAM Journal on Scientific Computing, Vol. 44, No. 1, 2022, pp. A554–A582. https://doi.org/10.1137/21M1411081

    work page doi:10.1137/21m1411081 2022

  7. [7]

    Reducedbasismethodsfortime-dependentproblems,

    Hesthaven, J.S., Pagliantini, C., andRozza, G., “Reducedbasismethodsfortime-dependentproblems,”ActaNumerica, Vol.31, 2022, pp. 265–345. https://doi.org/10.1017/S0962492922000058

    work page doi:10.1017/s0962492922000058 2022

  8. [8]

    Breaking the Kolmogorov barrier with nonlinear model reduction,

    Peherstorfer, B., “Breaking the Kolmogorov barrier with nonlinear model reduction,”Notices of the American Mathematical Society, Vol. 69, No. 5, 2022, pp. 725–733. https://doi.org/10.1090/noti2475

    work page doi:10.1090/noti2475 2022

  9. [9]

    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. https://doi.org/10.1016/j.jcp.2022. 111348

    work page doi:10.1016/j.jcp.2022 2022

  10. [10]

    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

    work page doi:10.1016/j.cma.2022 2023

  11. [11]

    Learning physics-based reduced-order models from data using nonlinear manifolds,

    Geelen, R., Balzano, L., Wright, S., and Willcox, K., “Learning physics-based reduced-order models from data using nonlinear manifolds,”Chaos: AnInterdisciplinaryJournalofNonlinearScience, Vol.34, No.3, 2024. https://doi.org/10.1063/5.0170105

    work page doi:10.1063/5.0170105 2024

  12. [12]

    Greedy construction of quadratic manifolds for nonlinear dimensionality reduction and nonlinear model reduction,

    Schwerdtner, P., and Peherstorfer, B., “Greedy construction of quadratic manifolds for nonlinear dimensionality reduction and nonlinear model reduction,” arXiv:2403.06732, 2024. https://doi.org/10.48550/arXiv.2403.06732

    work page doi:10.48550/arxiv.2403.06732 2024

  13. [13]

    Kernel principal component analysis,

    Schölkopf, B., Smola, A., and Müller, K.-R., “Kernel principal component analysis,”International Conference on Artificial Neural Networks, 1997, pp. 583–588. https://doi.org/10.1007/BFb0020217

    work page doi:10.1007/bfb0020217 1997

  14. [14]

    Kernel manifolds: Nonlinear-augmentation dimensionality reduction using reproducing kernel Hilbert spaces,

    Diaz, A. N., Needels, J. T., Tezaur, I. K., and Blonigan, P. J., “Kernel manifolds: Nonlinear-augmentation dimensionality reduction using reproducing kernel Hilbert spaces,”International Journal for Numerical Methods in Engineering, Vol. 126, No. 24, 2025, p. e70230. https://doi.org/https://doi.org/10.1002/nme.70230

    work page doi:10.1002/nme.70230 2025

  15. [15]

    LaSDI: Parametric latent space dynamics identification,

    Fries, W. D., He, X., and Choi, Y., “LaSDI: Parametric latent space dynamics identification,”Computer Methods in Applied Mechanics and Engineering, Vol. 399, 2022, p. 115436. https://doi.org/10.1016/j.cma.2022.115436

    work page doi:10.1016/j.cma.2022.115436 2022

  16. [16]

    Nonlinear model reduction for transport-dominated problems,

    Hesthaven, J. S., Peherstorfer, B., and Unger, B., “Nonlinear model reduction for transport-dominated problems,” arXiv:2602.01397, 2026. https://doi.org/10.48550/arXiv.2602.01397

    work page doi:10.48550/arxiv.2602.01397 2026

  17. [17]

    Moving finite elements. I,

    Miller, K., and Miller, R. N., “Moving finite elements. I,”SIAM Journal on Numerical Analysis, Vol. 18, No. 6, 1981, pp. 1019–1032. https://doi.org/10.1137/0718070

    work page doi:10.1137/0718070 1981

  18. [18]

    Projection approximation subspace tracking,

    Yang, B., “Projection approximation subspace tracking,”IEEE Transactions on Signal Processing, Vol. 43, No. 1, 1995, pp. 95–107. https://doi.org/10.1109/78.365290. 16

    work page doi:10.1109/78.365290 1995

  19. [19]

    Dynamical low-rank approximation,

    Koch, O., and Lubich, C., “Dynamical low-rank approximation,”SIAM Journal on Matrix Analysis and Applications, Vol. 29, No. 2, 2007, pp. 434–454. https://doi.org/10.1137/050639703

    work page doi:10.1137/050639703 2007

  20. [20]

    Dynamically orthogonal field equations for continuous stochastic dynamical systems,

    Sapsis, T. P., and Lermusiaux, P. F., “Dynamically orthogonal field equations for continuous stochastic dynamical systems,” Physica D: Nonlinear Phenomena, Vol. 238, No. 23-24, 2009, pp. 2347–2360. https://doi.org/10.1016/j.physd.2009.09.017

    work page doi:10.1016/j.physd.2009.09.017 2009

  21. [21]

    A minimization principle for the description of modes associated with finite-time instabilities,

    Babaee, H., and Sapsis, T. P., “A minimization principle for the description of modes associated with finite-time instabilities,” Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, Vol. 472, No. 2186, 2016, p. 20150779. https://doi.org/10.1098/rspa.2015.0779

    work page doi:10.1098/rspa.2015.0779 2016

  22. [22]

    Real-time reduced-order modeling of stochastic partial differential equations via time-dependent subspaces,

    Patil, P., and Babaee, H., “Real-time reduced-order modeling of stochastic partial differential equations via time-dependent subspaces,”Journal of Computational Physics, Vol. 415, 2020, p. 109511. https://doi.org/10.1016/j.jcp.2020.109511

    work page doi:10.1016/j.jcp.2020.109511 2020

  23. [23]

    On-the-fly reduced order modeling of passive and reactive species via time-dependent manifolds,

    Ramezanian, D., Nouri, A. G., and Babaee, H., “On-the-fly reduced order modeling of passive and reactive species via time-dependent manifolds,”Computer Methods in Applied Mechanics and Engineering, Vol. 382, 2021, p. 113882. https://doi.org/10.1016/j.cma.2021.113882

    work page doi:10.1016/j.cma.2021.113882 2021

  24. [24]

    Advection modes by optimal mass transfer,

    Iollo, A., and Lombardi, D., “Advection modes by optimal mass transfer,”Physical Review E, Vol. 89, No. 2, 2014, p. 022923. https://doi.org/10.1103/PhysRevE.89.022923

    work page doi:10.1103/physreve.89.022923 2014

  25. [25]

    Approximated Lax pairs for the reduced order integration of nonlinear evolution equations,

    Gerbeau, J.-F., and Lombardi, D., “Approximated Lax pairs for the reduced order integration of nonlinear evolution equations,” Journal of Computational Physics, Vol. 265, 2014, pp. 246–269. https://doi.org/10.1016/j.jcp.2014.01.047

    work page doi:10.1016/j.jcp.2014.01.047 2014

  26. [26]

    Manifold approximations via transported subspaces: Model reduction for transport-dominated problems,

    Rim, D., Peherstorfer, B., and Mandli, K. T., “Manifold approximations via transported subspaces: Model reduction for transport-dominated problems,”SIAM Journal on Scientific Computing, Vol. 45, No. 1, 2023, pp. A170–A199. https://doi.org/10.1137/20M1316998

    work page doi:10.1137/20m1316998 2023

  27. [27]

    Projection-based model reduction with dynamically transformed modes,

    Black, F., Schulze, P., and Unger, B., “Projection-based model reduction with dynamically transformed modes,”ESAIM: Mathematical Modelling and Numerical Analysis, Vol. 54, No. 6, 2020, pp. 2011–2043. https://doi.org/10.1051/m2an/2020046

    work page doi:10.1051/m2an/2020046 2020

  28. [28]

    ReducedordermodelsbasedonlocalPODplusGalerkinprojection,

    Rapún,M.-L.,andVega,J.M.,“ReducedordermodelsbasedonlocalPODplusGalerkinprojection,”JournalofComputational Physics, Vol. 229, No. 8, 2010, pp. 3046–3063. https://doi.org/10.1016/j.jcp.2009.12.029

    work page doi:10.1016/j.jcp.2009.12.029 2010

  29. [29]

    Dynamic data-driven reduced-order models,

    Peherstorfer, B., and Willcox, K., “Dynamic data-driven reduced-order models,”Computer Methods in Applied Mechanics and Engineering, Vol. 291, 2015, pp. 21–41. https://doi.org/10.1016/j.cma.2015.03.018

    work page doi:10.1016/j.cma.2015.03.018 2015

  30. [30]

    Online adaptive model reduction for nonlinear systems via low-rank updates,

    Peherstorfer, B., and Willcox, K., “Online adaptive model reduction for nonlinear systems via low-rank updates,”SIAM Journal on Scientific Computing, Vol. 37, No. 4, 2015, pp. A2123–A2150. https://doi.org/10.1137/140989169

    work page doi:10.1137/140989169 2015

  31. [31]

    Dynamical model reduction method for solving parameter-dependent dynamical systems,

    Billaud-Friess, M., and Nouy, A., “Dynamical model reduction method for solving parameter-dependent dynamical systems,” SIAM Journal on Scientific Computing, Vol. 39, No. 4, 2017, pp. A1766–A1792. https://doi.org/10.1137/16M1071493

    work page doi:10.1137/16m1071493 2017

  32. [32]

    The geometry of algorithms with orthogonality constraints,

    Edelman, A., Arias, T. A., and Smith, S. T., “The geometry of algorithms with orthogonality constraints,”SIAM Journal on Matrix Analysis and Applications, Vol. 20, No. 2, 1998, pp. 303–353. https://doi.org/10.1137/S0895479895290954

    work page doi:10.1137/s0895479895290954 1998

  33. [33]

    Riemannian geometry of Grassmann manifolds with a view on algorithmic computation,

    Absil, P.-A., Mahony, R., and Sepulchre, R., “Riemannian geometry of Grassmann manifolds with a view on algorithmic computation,”ActaApplicandaeMathematica,Vol.80,No.2,2004,pp.199–220. https://doi.org/10.1023/B:ACAP.0000013855. 14971.91

    work page doi:10.1023/b:acap.0000013855 2004

  34. [34]

    Dynamic subspace estimation with Grassmannian geodesics,

    Blocker, C. J., Raja, H., Fessler, J. A., and Balzano, L., “Dynamic subspace estimation with Grassmannian geodesics,” arXiv:2303.14851, 2023. https://doi.org/10.48550/arXiv.2303.14851

    work page doi:10.48550/arxiv.2303.14851 2023

  35. [35]

    Pymanopt: A Python toolbox for optimization on manifolds using automatic differentiation,

    Townsend, J., Koep, N., and Weichwald, S., “Pymanopt: A Python toolbox for optimization on manifolds using automatic differentiation,”Journal of Machine Learning Research, Vol. 17, No. 137, 2016, pp. 1–5. URL http://jmlr.org/papers/v17/16- 177.html

    2016

  36. [36]

    JAX: Composable transformations of Python+NumPy programs,

    Bradbury,J.,Frostig,R.,Hawkins,P.,Johnson,M.J.,Katariya,Y.,Leary,C.,Maclaurin,D.,Necula,G.,Paszke,A.,VanderPlas, J., Wanderman-Milne, S., and Zhang, Q., “JAX: Composable transformations of Python+NumPy programs,” GitHub:jax-ml/jax v0.3.13, 2018. URL http://github.com/jax-ml/jax

    2018

  37. [37]

    Turbulent airfoil wake large eddy simulation [Data set], University of Michigan - Deep Blue Data,

    Towne, A., Yeh, C., Patel, H., and Taira, K., “Turbulent airfoil wake large eddy simulation [Data set], University of Michigan - Deep Blue Data,” http://deepblue.lib.umich.edu/data/collections/kk91fk98z, 2022. 17

    2022

  38. [38]

    A database for reduced-complexity modeling of fluid flows,

    Towne, A., Dawson, S. T., Brès, G. A., Lozano-Durán, A., Saxton-Fox, T., Parthasarathy, A., Jones, A. R., Biler, H., Yeh, C.-A., Patel, H. D., et al., “A database for reduced-complexity modeling of fluid flows,”AIAA Journal, Vol. 61, No. 7, 2023, pp. 2867–2892. https://doi.org/10.2514/1.J062203

    work page doi:10.2514/1.j062203 2023

  39. [39]

    Resolvent-analysis-based design of airfoil separation control,

    Yeh, C.-A., and Taira, K., “Resolvent-analysis-based design of airfoil separation control,”Journal of Fluid Mechanics, Vol. 867, 2019, pp. 572–610. https://doi.org/10.1017/jfm.2019.163. 18

    work page doi:10.1017/jfm.2019.163 2019