Pith. sign in

REVIEW 3 major objections 6 minor 51 references

ResKoopNet: Learning Koopman Representations for Complex Dynamics with Spectral Residuals

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read ResKoopNet claims that minimizing the spectral residual over neural-network-learned observables recovers a more complete Koopman spectrum — discrete eigenvalues and continuous bands — than the filtering approach of ResDMD, using far fewer…

desk verdict Novel residual-based dictionary learning for Koopman spectra, but the theoretical guarantees are unsupported and the closed-form update is just EDMD. read the letter →

arxiv 2501.00701 v4 pith:JZHXKSAF submitted 2025-01-01 cs.LG math.DS

classification cs.LGmath.DS MSC 37M1065P9968T07
keywords Koopmanoperatorspectralresidualinclusionproblemneuralnetworkdictionarylearningcontinuousspectrumpseudospectrumdynamicmodedecompositionhigh-dimensionaldynamicalsystems
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 sets out to cure a known failure mode of data-driven Koopman spectral analysis: the 'spectral inclusion' problem, where a computed spectrum is confined to a subset of the true spectrum and can simply miss parts of it. Its proposed method, ResKoopNet, treats the spectral residual — a data-estimated measure of how far a candidate eigenvalue–eigenfunction pair is from satisfying the Koopman eigenvalue equation — as the objective to minimize, rather than as a post-hoc filter for rejecting polluted eigenvalues as ResDMD does. A neural network learns the dictionary of observables while the Koopman matrix is updated in closed form at each step, so the optimization searches for genuinely accurate eigenpairs instead of refining a precomputed set. The payoff, if the claims hold, is practical: the full pendulum spectrum is recovered with roughly 300 observables where ResDMD needs hundreds more, the dominant pressure field of a turbulent airfoil flow emerges as the lowest-residual Koopman mode, and eigenfunctions from mouse visual-cortex recordings separate neatly by stimulus identity where three comparison methods fail.

What carries the argument

The load-bearing object is the empirical spectral residual $\text{cres}(\lambda, \phi)^2 = (1/m)\, v^*[\Psi_Y^* \Psi_Y - \lambda(\Psi_X^* \Psi_Y)^* - \bar\lambda \Psi_X^* \Psi_Y + |\lambda|^2 \Psi_X^* \Psi_X]\, v$, which estimates, from snapshots alone, how far the pair $(\lambda, \phi = \Psi v)$ deviates from the Koopman eigenvalue equation $K\phi = \lambda\phi$. The paper's move is to sum this residual over all eigenpairs and minimize the sum as a loss — equivalently $\|(\Psi_Y - \Psi_X K)V\|_F^2$ — while a feedforward network re-shapes the dictionary $\Psi(x; \theta)$. The Koopman matrix is updated by the closed form $K = G^\dagger A$ at each step, and the pseudospectrum, which captures continuous spectral regions, is assembled by scanning candidate complex frequencies and recording where the minimal residual falls below a threshold.

What would settle it

Test the optimality claim on a tiny system with a known Koopman spectrum: compute the spectrum from ResKoopNet's fixed-$V$ closed form $K = G^\dagger A$ and compare it with the spectrum of a $K$ obtained by minimizing the same loss jointly over $K$ and its eigenvector matrix $V$, with no fixed-$V$ shortcut; if the two spectra differ materially, the closed form does not minimize the residual and the theoretical claim collapses. A complementary check is a random-seed sweep of the pendulum experiment at the claimed $N_K = 300$: the paper's own appendix shows the method is unstable with one or two hidden layers, so repeated runs that fail to place eigenvalues on the unit circle would falsify the efficiency claim.

Watch

Extended reading notes

Core claim

On its own terms, ResKoopNet establishes that the Koopman spectrum can be learned directly by minimizing the total spectral residual $J = (1/m)\|(\Psi_Y - \Psi_X K)V\|_F^2$ over all computed eigenpairs, where $V$ is the eigenvector matrix of the Koopman matrix $K$ and $\Psi$ is a dictionary of observables parameterized by a neural network. For a fixed dictionary the paper derives a closed-form optimal Koopman matrix, $K = G^\dagger A$ with $G = (1/m)\Psi_X^* \Psi_X$ and $A = (1/m)\Psi_X^* \Psi_Y$, and alternates this update with gradient descent on the network parameters until $J(\theta)$ falls below a threshold. Because a vanishing spectral residual implies $\|K\phi - \lambda\phi\|$ is small, the computed eigenpairs inherit the convergence guarantees of the residual-based spectral theory: as data and dictionary size grow, eigenpairs and pseudospectra converge to the operator's true discrete and continuous spectra. The central discovery is therefore that spectral accuracy, not predictive accuracy, should be the training objective, and that this objective can be optimized through a dictionary learned from data rather than fixed in advance.

Load-bearing premise

The load-bearing premise is that the closed-form Koopman matrix $K = G^\dagger A$ truly minimizes the spectral-residual loss, even though the derivation differentiates the loss while holding the eigenvector matrix $V$ fixed even though $V$ is by definition the eigenvector matrix of $K$; if that optimality claim fails, the theoretical guarantees reduce to those of EDMD with a residual-weighted dictionary.

Editorial extensions

If this is right

  • If the central claim holds, the spectral inclusion problem is resolved: optimizing eigenpairs directly against the spectral residual recovers discrete eigenvalues and continuous spectral bands that ResDMD's filter-only approach can miss, including spectra that are otherwise trivial.
  • Koopman-mode extraction improves on real high-dimensional data: the lowest-residual mode in the turbulent airfoil example reconstructs the dominant pressure field, a global spatial structure that kernel-based ResDMD and Hankel-DMD do not recover from the same dataset.
  • Substantially smaller dictionaries suffice for accurate spectra: roughly 300–350 learned observables approximate the pendulum's full unit-circle spectrum, compared with roughly 460–964 fixed basis functions for ResDMD.
  • Koopman eigenfunctions learned this way separate latent dynamical states: on five mouse visual-cortex recordings, eigenfunctions cluster cleanly by the six video stimuli (low Davies-Bouldin index), while Hankel-DMD, EDMD with RBF bases, and kernel ResDMD show no comparable separation.

Reading between the lines

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

  • The paper's guarantee chain, read closely, rests on the ResDMD residual-to-spectrum theorem plus a Barron-space approximation argument; the closed-form $K = G^\dagger A$ update is what a skeptical reader would test first, since the derivation holds $V$ fixed even though $V$ is defined by $K$.
  • Because the objective is purely spectral, nothing in the loss rewards prediction of future states; a natural extension is adding a small prediction-error term to test whether spectral accuracy and forecasting accuracy are compatible at fixed dictionary size, and whether spectral training improves long-horizon forecasts on chaotic systems.
  • The continuous-spectrum claim is demonstrated through grid-based pseudospectra; an independent check would compare ResKoopNet's $\varepsilon$-pseudospectrum with an analytically known continuous spectrum — for instance a rotation system — to verify that the recovered band has the right location and width.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper introduces ResKoopNet, a method that trains a neural-network parameterized dictionary by minimizing the empirical spectral residual (Eq. 3.3) over computed Koopman eigenpairs, with the aim of addressing the spectral inclusion problem and recovering both discrete and continuous spectra. The authors propose an alternating scheme: update the dictionary parameters θ by gradient descent on the residual loss J(θ), and at each step set the Koopman matrix to K(θ) = (G(θ)+σI)^{-1}A(θ), which they claim minimizes J. The method is tested on a pendulum, a turbulent airfoil flow, and calcium-imaging data from mouse visual cortex, where it is compared with EDMD, EDMD-DL, Hankel-DMD, and ResDMD. The paper also includes a convergence discussion based on Barron-space approximation theory and an SGD rate argument in Appendix A.4.

Significance. If the theoretical claims were established, ResKoopNet would be a meaningful contribution: it would offer a principled way to learn dictionaries directly for spectral accuracy, potentially reducing the number of observables needed for continuous spectra and scaling to high-dimensional systems. The empirical results, particularly the pendulum spectrum with NK=300 and the recovery of the pressure field in the turbulence example, are suggestive and the work addresses a real limitation of ResDMD, namely that filtering precomputed spectra cannot discover missing spectral components. However, the advertised theoretical guarantees are not currently supported by the analysis in the manuscript. The derivation of the closed-form Koopman matrix has a load-bearing gap, and the convergence argument relies on assumptions that the paper itself acknowledges are not satisfied. The empirical evaluation is therefore the main strength of the paper, but the central theoretical claims need substantial revision before publication.

major comments (3)
  1. [Section 3.2, Eqs. (3.5)-(3.6) and Appendix A.3] The derivation of the closed-form optimal Koopman matrix K = G†A differentiates the loss J = (1/m)||(ΨY − ΨX K)V||_F^2 with respect to K while treating the eigenvector matrix V as a constant. However, V is defined as the eigenvector matrix of K, so V depends on K. The calculation in Appendix A.3 omits the terms arising from ∂V/∂K. Consequently, the paper does not establish that K = G†A minimizes J over all K. At best, the calculation shows that K = G†A minimizes the residual for a fixed V, which is not the objective being optimized. This is a load-bearing issue because the method's theoretical foundation rests on the claim that Eq. (3.6) is the optimal Koopman matrix for the spectral-residual loss. The authors should either provide a rigorous derivation that accounts for the V dependence, or explicitly state that Eq. (3.6) is a heuristic choice and analyze the actual objective being minimized by the algorithm.
  2. [Appendix A.4, Assumption A.4(b) and the SGD convergence argument] The convergence analysis assumes that J(θ) is strongly convex and Lipschitz continuous in θ, and uses this assumption to invoke an O(1/n) SGD rate and to conclude that J(θ_n) tends to zero. In the very next paragraph, the paper acknowledges that 'In practice, J(θ) is non-convex due to the neural network.' Strong convexity is therefore not satisfied, and the stated SGD convergence result does not apply. The claim that J(θ_n) → 0 as n and m tend to infinity is thus unsupported. The authors need to replace this argument with a non-convex convergence analysis, or significantly weaken the theoretical claims to state what is actually proven.
  3. [Appendix A.4, paragraph on spectral convergence and data-dependent dictionaries] The paper invokes ResDMD's convergence results, e.g., Colbrook and Townsend [8, Theorem B.1, Lemma B.1], to argue that J(θ) → 0 implies convergence of the computed eigenpairs and pseudospectrum to the true Koopman spectrum. Those results are for a fixed dictionary, whereas ResKoopNet uses a dictionary Ψ(x; θ) that is trained on the same snapshots used to define the empirical residual. No sample-splitting, regularization, or generalization bound is provided to show that minimizing the training residual controls the true spectral residual. The sentence claiming that uniform convergence follows from density of B_NK and Dini's theorem is not justified for data-dependent dictionaries. A theorem with explicit assumptions relating the number of snapshots m, the dictionary size NK, and the network capacity to the spectral error is needed to substantiate the abstract's claim of theoretical guarantees.
minor comments (6)
  1. [Abstract and Section 1] The abstract states that the approach 'provides theoretical guarantees while maintaining computational adaptability,' but no formal theorem for ResKoopNet is stated in the main text; the convergence discussion in Appendix A.4 is conditional on assumptions that are acknowledged to be violated. The wording should be tempered to match the actual results.
  2. [Section 3.2, Eq. (3.6) and Algorithm 1] Eq. (3.6) presents the optimal Koopman matrix as K = G†A, but Algorithm 1 and Remark 3.2 use the regularized form K = (G+σI)^{-1}A. The relationship between these two expressions and the role of σ in the theory should be clarified.
  3. [Appendix A.1] The statement 'the source code will be available at this link' does not include an actual URL. For a reproducibility-focused journal, a working link should be provided.
  4. [Section 4.3] The paper says 'we trained dictionaries on all snapshots from each mouse to avoid overfitting,' but this means the evaluation of clustering is performed on the same data used for training. This should be stated explicitly as an in-sample evaluation, and the possibility of overfitting to trial identities should be discussed.
  5. [Section 4.2 and Appendix A.8.2] The Hankel-DMD comparison in the turbulence example is described as having small spectral residuals, yet the paper argues that the residual metric 'does not fully extend to the Hankel-DMD setting.' If the residual is not a valid comparison for Hankel-DMD, the quantitative comparison in Figure 9 and the associated discussion should be framed more carefully.
  6. [Appendix A.4, Assumption A.4(a)] Assumption A.4(a) assumes that there is a finite-dimensional invariant subspace B_NK spanned by optimal dictionary functions. For systems with continuous spectra, such a finite-dimensional exactly invariant subspace generally does not exist; this assumption should be stated and discussed, since it is not satisfied by the pendulum example with continuous spectrum.

Circularity Check

1 steps flagged · score 4.0 of 10

The closed-form Koopman matrix K = G†A is, for fixed dictionary, the EDMD solution by construction; the paper's claim that it is derived from spectral-residual rather than prediction-error minimization is a renaming of a known result. The dictionary-learning loss and empirical benchmarks remain independent, so circularity is partial.

  1. renaming known result [Section 3.2, Eqs. (3.5)–(3.6) and the paragraph following Remark 3.2]
    "minimizing J is also equivalent to the following minimization problem: min_˜K J = min_˜K 1/m ∥(ΨY − ΨX ˜K)V ∥²_F , (3.5) where each column of the matrix V is an (right) eigenvector vi of the matrix ˜K. Thus, with dictionary Ψ fixed, we can obtain a closed form for the optimal Koopman matrix ˜K as following ˜K = G†A, (3.6)... While they share a similar matrix expression, this expression is derived from minimizing the spectral residual in Eq. (3.3) rather than the prediction error that EDMD minimizes."

    For fixed Ψ, V is the eigenvector matrix of K and is invertible. The map K ↦ KV is a bijection, so min_K ∥(ΨY − ΨX K)V∥_F is the same least-squares problem as min_K ∥ΨY − ΨX K∥_F; both have solution K = Ψ_X† Ψ_Y = G†A. Thus Eq. (3.6) is exactly the EDMD Koopman matrix by construction, and the asserted distinction between 'derived from minimizing the spectral residual' and 'prediction error that EDMD minimizes' is not realized for the closed-form K. The claimed theoretical foundation for the Koopman matrix therefore reduces to renaming a known result; only the data-dependent dictionary loss J(θ) provides independent content.

full rationale

The paper's main new algorithmic content is the dictionary-learning loop that minimizes the empirical spectral residual J(θ) over neural-network parameters, and the experimental comparisons use external ground truth (unit-circle spectrum, airfoil pressure field, visual-stimulus trial labels), so the overall method is not self-validating. However, one load-bearing step in the derivation chain is circular in the sense of renaming a known result: the closed-form optimal Koopman matrix in Eq. (3.6) coincides with EDMD's K = G†A because the eigenvector matrix V is invertible, making the projected residual minimization equivalent to EDMD's prediction-error minimization for a fixed dictionary. The paper's assertion that the shared expression arises from a fundamentally different theoretical foundation is therefore not supported by its own equations. The residual is also used both as the training loss and as the mode-ordering/evaluation criterion, but since the headline claims are checked against ground-truth physical and biological patterns, this is a metric-alignment caveat rather than a circular prediction. The self-citation [46] is not load-bearing. Appendix A.4's admission that J(θ) is non-convex despite Assumption A.4(b) using strong convexity is a gap in the claimed theoretical guarantees, but it is a correctness risk, not a circularity. Overall, the central contribution retains independent empirical content, so the paper is only partially circular.

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

No new physical entities. The main burden is mathematical: two unjustified optimality and convexity assumptions, plus user-chosen hyperparameters.

free parameters (5)
  • Number of dictionary observables NK = 25-350 depending on experiment
    Chosen per experiment by hand; comparative claims (e.g., NK=300 vs 964) depend on it.
  • Neural network architecture (hidden layers, neurons) = e.g., 3x300, 3x200
    Selected via hyperparameter scans (Appendix A.7.2, Fig. 17); performance varies with architecture.
  • Regularization parameter sigma and loss threshold epsilon = not specified numerically
    Algorithm 1 inputs; no values reported, affecting reproducibility.
  • SVD truncation ranks = 150 (turbulence), 24/300 (neural)
    Preprocessing choice; the learned dictionary is built on the truncated coordinates.
  • Hankel-DMD time delay (baseline) = 150, 50, or 5 depending on experiment
    Baseline comparison parameter; different delay per experiment complicates comparisons.
assumptions (5)
  • standard math The Koopman operator K on L2(Omega, mu) has a spectral decomposition into point and continuous spectra (Section 2).
    Standard Koopman theory; the paper relies on KMD expansion.
  • domain assumption The optimal dictionary functions spanning the Koopman invariant subspace lie in Barron space (Assumption A.4(a)).
    Needed for the O(1/sqrt(r)) neural approximation rate; not verified for the systems tested.
  • ad hoc to paper The loss J(theta) is strongly convex and Lipschitz continuous in theta (Assumption A.4(b)).
    Used to claim SGD convergence O(1/n); the paper admits J is non-convex for neural networks, so this assumption is false.
  • ad hoc to paper K = G†A minimizes the spectral-residual loss J even though the eigenvector matrix V depends on K.
    Unproved; the derivation in Appendix A.3 holds V fixed. Load-bearing for the theoretical claim.
  • standard math Empirical Gram matrices converge to L2 inner products as m goes to infinity (Eq. 3.2).
    Law of large numbers; standard in EDMD and ResDMD.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ResKoopNet: Learning Koopman Representations for Complex Dynamics with Spectral Residuals." pith.science (2026). https://pith.science/paper/JZHXKSAF

@misc{pith2026250100701,
  author       = {Pith},
  title        = {Pith review of: ResKoopNet: Learning Koopman Representations for Complex Dynamics with Spectral Residuals},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JZHXKSAF}},
  note         = {Machine review of arXiv:2501.00701}
}
read the original abstract

Analyzing the long-term behavior of high-dimensional nonlinear dynamical systems remains a significant challenge. While the Koopman operator framework provides a powerful global linearization tool, current methods for approximating its spectral components often face theoretical limitations and depend on predefined dictionaries. Residual Dynamic Mode Decomposition (ResDMD) advanced the field by introducing the \emph{spectral residual} to assess Koopman operator approximation accuracy; however, its approach of only filtering precomputed spectra prevents the discovery of the operator's complete spectral information, a limitation known as the `spectral inclusion' problem. We introduce ResKoopNet (Residual-based Koopman-learning Network), a novel method that directly addresses this by explicitly minimizing the \emph{spectral residual} to compute Koopman eigenpairs. This enables the identification of a more precise and complete Koopman operator spectrum. Using neural networks, our approach provides theoretical guarantees while maintaining computational adaptability. Experiments on a variety of physical and biological systems show that ResKoopNet achieves more accurate spectral approximations than existing methods, particularly for high-dimensional systems and those with continuous spectra, which demonstrates its effectiveness as a tool for analyzing complex dynamical systems.

Figures

Figures reproduced from arXiv: 2501.00701 by the authors.

Figure 1
Figure 1. The optimizer on the right searches for the optimal dictionary by minimizing Eq. (3.3). [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The four plots depict the spectrum of the Koop [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 4
Figure 4. Comparison with classical methods. Eigenvalue spectra computed from a [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figures from the paper (13 more)
Figure 5
Figure 5. Figure 5: Turbulence detection using Koopman modes from 250 observables. (a) shows the original 2D pressure field. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: ResKoopNet outperforms Hankel-DMD in identifying latent dynamic structures in neural signals with a [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Comparison across different dictionary sizes by ResDMD. [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]
Figure 8
Figure 8. Figure 8: Comparisons of neural network architectures. Each row corresponds to the number of hidden layers (from 1 [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: The plots illustrate turbulence detection using the four Koopman modes computed by Hankel-DMD, which [PITH_FULL_IMAGE:figures/full_fig_p021_9.png]
Figure 10
Figure 10. Figure 10: Trial-averaged neural responses to 6 video stimuli (marked by 6 states). [PITH_FULL_IMAGE:figures/full_fig_p023_10.png]
Figure 11
Figure 11. Figure 11: (A) Mean firing rates (calcium spike) of each mouse in response to 6 video stimuli (marked as states). [PITH_FULL_IMAGE:figures/full_fig_p024_11.png]
Figure 12
Figure 12. Figure 12: Comparison of clustering quality based on Davies-Bouldin Indices for clustering with mean firing rates [PITH_FULL_IMAGE:figures/full_fig_p025_12.png]
Figure 13
Figure 13. Figure 13: State Partition performance with eigenfunctions for ResKoopNet (50 bases) and Hankel-DMD in 2D space [PITH_FULL_IMAGE:figures/full_fig_p025_13.png]
Figure 14
Figure 14. Figure 14: Full results of EDMD with RBF basis. (A) 1301 Koopman eigenfunctions estimated by EDMD with RBF [PITH_FULL_IMAGE:figures/full_fig_p026_14.png]
Figure 15
Figure 15. Figure 15: Same as Figure 14 but estimated with Kernel ResDMD, with 299 basis of the Koopman subspace, thus 299 [PITH_FULL_IMAGE:figures/full_fig_p027_15.png]
Figure 16
Figure 16. Figure 16: Davies-Bouldin Indices evaluating the clustering performance of dynamical components learned by four [PITH_FULL_IMAGE:figures/full_fig_p027_16.png]
Figure 17
Figure 17. Figure 17: Davies-Bouldin Indices for different neural network hyperparameters in training ResKoopNet with 50 bases [PITH_FULL_IMAGE:figures/full_fig_p028_17.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 34 canonical work pages

  1. [1]

    Ergodic theory, dynamic mode decomposition, and computation of spectral properties of the koopman operator

    Hassan Arbabi and Igor Mezic. Ergodic theory, dynamic mode decomposition, and computation of spectral properties of the koopman operator. SIAM Journal on Applied Dynamical Systems, 16(4):2096–2126, 2017

  2. [2]

    A.R. Barron. Universal approximation bounds for superpositions of a sigmoidal function. IEEE Transactions on Information Theory, 39(3):930–945, 1993. doi: 10.1109/18.256500

  3. [3]

    Mapping multiple features in the population response of visual cortex

    Amit Basole, Leonard E White, and David Fitzpatrick. Mapping multiple features in the population response of visual cortex. Nature, 423(6943):986–990, 2003

  4. [4]

    Optimization methods for large-scale machine learning

    Léon Bottou, Frank E Curtis, and Jorge Nocedal. Optimization methods for large-scale machine learning. SIAM review, 60(2):223–311, 2018

  5. [5]

    Boyd.Chebyshev and Fourier Spectral Methods: Second Revised Edition

    J.P. Boyd.Chebyshev and Fourier Spectral Methods: Second Revised Edition. Dover Books on Mathematics. Dover Publications, 2013. ISBN 9780486141923. URL https://books.google.ca/books?id=b4TCAgAAQBAJ

  6. [6]

    Brunton and J

    Steven L. Brunton and J. Nathan Kutz. Data-Driven Science and Engineering: Machine Learning, Dynamical Systems, and Control. Cambridge University Press, 2019

  7. [7]

    Chaos as an intermittently forced linear system

    Steven L Brunton, Bingni W Brunton, Joshua L Proctor, Eurika Kaiser, and J Nathan Kutz. Chaos as an intermittently forced linear system. Nature communications, 8(1):19, 2017

  8. [8]

    Rigorous data-driven computation of spectral properties of koopman operators for dynamical systems

    Matthew J Colbrook and Alex Townsend. Rigorous data-driven computation of spectral properties of koopman operators for dynamical systems. Communications on Pure and Applied Mathematics, 77(1):221–283, 2024

Show all 51 references
  1. [9]

    Residual dynamic mode decomposition: robust and verified koopmanism

    Matthew J Colbrook, Lorna J Ayton, and Máté Sz˝oke. Residual dynamic mode decomposition: robust and verified koopmanism. Journal of Fluid Mechanics, 955:A21, 2023

  2. [10]

    G. Cybenko. Approximation by superpositions of a sigmoidal function. Mathematics of Control, Signals, and Systems , 2(4):303–314, December 1989. doi: 10.1007/BF02551274. URL https://hal.science/ hal-03753170

  3. [11]

    David M. Grobman. Homeomorphisms of systems of differential equations. Doklady Akademii Nauk SSSR, 128 (5):880–881, 1959

  4. [12]

    A lemma in the theory of structural stability of differential equations

    Philip Hartman. A lemma in the theory of structural stability of differential equations. Proceedings of the American Mathematical Society, 11(4):610–620, 1960. 11

  5. [13]

    S.S. Haykin. Neural Networks and Learning Machines. Pearson International Edition. Pearson, 2009. ISBN 9780131293762. URL https://books.google.ca/books?id=KCwWOAAACAAJ

  6. [14]

    Primary visual cortex straightens natural video trajectories

    Olivier J Hénaff, Yoon Bai, Julie A Charlton, Ian Nauhaus, Eero P Simoncelli, and Robbe LT Goris. Primary visual cortex straightens natural video trajectories. Nature communications, 12(1):5982, 2021

  7. [15]

    Koopman operators with intrinsic observables in rigged reproducing kernel hilbert spaces, 2024

    Isao Ishikawa, Yuka Hashimoto, Masahiro Ikeda, and Yoshinobu Kawahara. Koopman operators with intrinsic observables in rigged reproducing kernel hilbert spaces, 2024

  8. [16]

    A kernel-based method for data-driven koopman spectral analysis

    I Kevrekidis, Clarence W Rowley, and M Williams. A kernel-based method for data-driven koopman spectral analysis. Journal of Computational Dynamics, 2(2):247–265, 2016

  9. [17]

    Hamiltonian systems and transformation in hilbert space

    Bernard O Koopman. Hamiltonian systems and transformation in hilbert space. Proceedings of the National Academy of Sciences, 17(5):315, 1931

  10. [18]

    Koopman and John von Neumann

    Bernard O. Koopman and John von Neumann. Dynamical systems of continuous spectra. Proceedings of the National Academy of Sciences, 18(3):255–263, 1932. doi: 10.1073/pnas.18.3.255

  11. [19]

    On convergence of extended dynamic mode decomposition to the koopman operator

    Milan Korda and Igor Mezi´c. On convergence of extended dynamic mode decomposition to the koopman operator. Journal of Nonlinear Science, 28:687–710, 2018

  12. [20]

    Nonmetric multidimensional scaling: a numerical method

    Joseph B Kruskal. Nonmetric multidimensional scaling: a numerical method. Psychometrika, 29(2):115–129, 1964

  13. [21]

    Linearization in the large of nonlinear systems and koopman operator spectrum

    Yueheng Lan and Igor Mezi´c. Linearization in the large of nonlinear systems and koopman operator spectrum. Physica D: Nonlinear Phenomena, 242(1):42–53, 2013

  14. [22]

    Extended dynamic mode decomposition with dictionary learning: A data-driven adaptive spectral decomposition of the koopman operator

    Qianxiao Li, Felix Dietrich, Erik M Bollt, and Ioannis G Kevrekidis. Extended dynamic mode decomposition with dictionary learning: A data-driven adaptive spectral decomposition of the koopman operator. Chaos: An Interdisciplinary Journal of Nonlinear Science, 27(10), 2017

  15. [23]

    Deep learning for universal linear embeddings of nonlinear dynamics

    Bethany Lusch, J Nathan Kutz, and Steven L Brunton. Deep learning for universal linear embeddings of nonlinear dynamics. Nature Communications, 9(1):4950, 2018

  16. [24]

    Vampnets for deep learning of molecular kinetics.Nature communications, 9(1):5, 2018

    Andreas Mardt, Luca Pasquali, Hao Wu, and Frank Noé. Vampnets for deep learning of molecular kinetics.Nature communications, 9(1):5, 2018

  17. [25]

    Umap: Uniform manifold approximation and projection for dimension reduction

    Leland McInnes, John Healy, and James Melville. Umap: Uniform manifold approximation and projection for dimension reduction. arXiv preprint arXiv:1802.03426, 2018

  18. [26]

    Spectral properties of dynamical systems, model reduction and decompositions

    Igor Mezi´c. Spectral properties of dynamical systems, model reduction and decompositions. Nonlinear Dynamics, 41(1-3):309–325, 2005

  19. [27]

    Analysis of fluid flows via spectral properties of the koopman operator

    Igor Mezi´c. Analysis of fluid flows via spectral properties of the koopman operator. Annual review of fluid mechanics, 45:357–378, 2013

  20. [28]

    Natural scene evoked population dynamics across cat primary visual cortex captured with voltage-sensitive dye imaging

    Selim Onat, Peter König, and Dirk Jancke. Natural scene evoked population dynamics across cat primary visual cortex captured with voltage-sensitive dye imaging. Cerebral cortex, 21(11):2542–2554, 2011

  21. [29]

    Linearly recurrent autoencoder networks for learning dynamics

    Samuel E Otto and Clarence W Rowley. Linearly recurrent autoencoder networks for learning dynamics. SIAM Journal on Applied Dynamical Systems, 18(1):558–593, 2019

  22. [30]

    Approximation theory of the mlp model in neural networks

    Allan Pinkus. Approximation theory of the mlp model in neural networks. Acta Numerica, 8:143–195, 1999. doi: 10.1017/S0962492900002919

  23. [31]

    Schetzen

    M. Schetzen. The Volterra and Wiener Theories of Nonlinear Systems. Krieger Pub., 2006. ISBN 9781575242835. URL https://books.google.ca/books?id=hgFVAAAAYAAJ

  24. [32]

    Dynamic mode decomposition and its variants

    Peter J Schmid. Dynamic mode decomposition and its variants. Annual Review of Fluid Mechanics , 54(1): 225–254, 2022

  25. [33]

    Slotine and W

    J.J.E. Slotine and W. Li. Applied Nonlinear Control . Prentice Hall, 1991. ISBN 9780130408907. URL https://books.google.ca/books?id=cwpRAAAAMAAJ

  26. [34]

    Learning koopman invariant subspaces for dynamic mode decomposition

    Naoya Takeishi, Yoshinobu Kawahara, and Takehisa Yairi. Learning koopman invariant subspaces for dynamic mode decomposition. Advances in neural information processing systems, 30, 2017

  27. [35]

    Detecting strange attractors in turbulence

    Floris Takens. Detecting strange attractors in turbulence. In Dynamical Systems and Turbulence, Warwick 1980: proceedings of a symposium held at the University of Warwick 1979/80, pages 366–381. Springer, 2006

  28. [36]

    Kernel embedding based variational approach for low-dimensional approximation of dynamical systems

    Wenchong Tian and Hao Wu. Kernel embedding based variational approach for low-dimensional approximation of dynamical systems. Computational Methods in Applied Mathematics, 21(3):635–659, 2021

  29. [37]

    Tu, Clarence W

    Jonathan H. Tu, Clarence W. Rowley, Dirk M. Luchtenburg, Steven L. Brunton, and J. Nathan Kutz. On dynamic mode decomposition: Theory and applications. Journal of Computational Dynamics, 1(2):391–421, 2014. 12

  30. [38]

    Retrospective for the dynamic sensorium competition for predicting large-scale mouse primary visual cortex activity from videos

    Polina Turishcheva, Paul Fahey, Michaela Vystrˇcilová, Laura Hansel, Rachel Froebe, Kayla Ponder, Yongrong Qiu, Konstantin Willeke, Mohammad Bashiri, Ruslan Baikulov, et al. Retrospective for the dynamic sensorium competition for predicting large-scale mouse primary visual cor...

  31. [39]

    The dynamic sensorium competition for predicting large-scale mouse visual cortex activity from videos

    Polina Turishcheva, Paul G Fahey, Michaela Vystrˇcilová, Laura Hansel, Rachel Froebe, Kayla Ponder, Yongrong Qiu, Konstantin F Willeke, Mohammad Bashiri, Eric Wang, et al. The dynamic sensorium competition for predicting large-scale mouse visual cortex activity from videos. Ar...

  32. [40]

    Visualizing data using t-sne

    Laurens Van der Maaten and Geoffrey Hinton. Visualizing data using t-sne. Journal of machine learning research, 9(11), 2008

  33. [41]

    Barron spaces and the compositional function spaces for neural network models

    E Weinan, Chao Ma, and Lei Wu. Barron spaces and the compositional function spaces for neural network models. arXiv preprint arXiv:1906.08039, 2019

  34. [42]

    Towards a mathematical understanding of neural network- based machine learning: what we know and what we don’t

    E Weinan, Chao Ma, Stephan Wojtowytsch, and Lei Wu. Towards a mathematical understanding of neural network- based machine learning: what we know and what we don’t. arxiv e-prints. arXiv preprint arXiv:2009.10713, 8, 2020

  35. [43]

    S. Wiggins. Introduction to Applied Nonlinear Dynamical Systems and Chaos . Texts in Applied Mathe- matics. Springer New York, 2003. ISBN 9780387001777. URL https://books.google.ca/books?id= RSI4RGdwnU4C

  36. [44]

    Data-driven approximation of the koopman operator: Extending dynamic mode decomposition

    Matthew O Williams, Ioannis G Kevrekidis, and Clarence W Rowley. Data-driven approximation of the koopman operator: Extending dynamic mode decomposition. Journal of Nonlinear Science, 25(6):1307–1346, 2015

  37. [45]

    Variational approach for learning markov processes from time series data

    Hao Wu and Frank Noé. Variational approach for learning markov processes from time series data. Journal of Nonlinear Science, 30(1):23–66, 2020

  38. [46]

    A data- driven framework for koopman semigroup estimation in stochastic dynamical systems, 2025

    Yuanchao Xu, Kaidi Shao, Isao Ishikawa, Yuka Hashimoto, Nikos Logothetis, and Zhongwei Shen. A data- driven framework for koopman semigroup estimation in stochastic dynamical systems, 2025. URL https: //arxiv.org/abs/2501.13301

  39. [47]

    two-layer neural network

    Jinhui Yang, Juan Zhao, Junqiang Song, Jianping Wu, Chengwu Zhao, and Hongze Leng. A hybrid method using havok analysis and machine learning for predicting chaotic time series. Entropy, 24(3), 2022. ISSN 1099-4300. doi: 10.3390/e24030408. URL https://www.mdpi.com/1099-4300/24/...

  40. [48]

    This choice ensures the dictionary size is the same as Hankel DMD, making it a fair comparison between the performance of these two methods

    For ResKoopNet, we selected 25 trained basis functions and 24 first-order monomial basis functions as the dictionary for the 24 reduced observables. This choice ensures the dictionary size is the same as Hankel DMD, making it a fair comparison between the performance of these ...

  41. [49]

    Therefore, it is impossible to choose the same dictionary size as the ResKoopNet example

    For Hankel DMD, the number of delays (as dictionary size/number of eigenfunctions) is first constrained by the temporal sample size (i.e., snapshot size) because it cannot exceed the maximum snapshot size. Therefore, it is impossible to choose the same dictionary size as the R...

  42. [50]

    For RBF basis, in principle, we can use the same dictionary size. However, our previous experience with a similar dataset and the results of using the RBF basis for the EDMD method all suggest that the performance will be better with more dictionary functions. Therefore, we ch...

  43. [51]

    Therefore, we cannot make the dictionary size consistent with the ResKoopNet example

    For Kernel ResDMD, the dictionary size is theoretically determined to be the number of snapshots. Therefore, we cannot make the dictionary size consistent with the ResKoopNet example. Based on the above justifications, we believe our choices of dictionary sizes are reasonable ...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.