Pith. sign in

REVIEW 3 major objections 5 minor 36 references

Positive-Allocation Companion Predictors for Nonlinear Dynamics and Their Finite-Difference Diagnostics

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Positive-allocation constraints force companion predictors to have spectra in the closed unit disk, making their model trajectories bounded and finite differences act as spectral filters.

desk verdict Sound elementary spectral theory and an honest experimental section, but the predictive transfer claim is not established; worth a referee's time as a modest methodological note. read the letter →

arxiv 2607.16529 v1 pith:KWPKQETA submitted 2026-07-17 math.DS cs.NAmath.NA

Cynthia V. Flores , James E. Pascoe This is my paper · ORCID
classification math.DScs.NAmath.NA MSC 37M1037N4065P9947A10
keywords positiveallocationcompanionmatrixnonlineardynamicsunit-diskcontainmentspectralfiltersfinite-differencediagnosticsdata-drivenpredictionconvexcombination
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

This paper introduces positive-allocation companion predictors for nonlinear dynamical systems. The method expresses the latest observed snapshot as a convex combination—nonnegative weights summing to one—of earlier snapshots, and encodes those weights as the last column of a companion matrix. The main theorem shows that this constraint forces all eigenvalues of the companion matrix into the closed unit disk and makes 1 an eigenvalue, so the generated model trajectory cannot grow exponentially. The paper then proves that finite differences of the model trajectory act as spectral filters, with first differences multiplying modes by (λ−1) and second differences by (λ−1)², and it derives non-modal bounds that avoid diagonalization. Numerical experiments on FitzHugh–Nagumo and SIR models show these diagnostics remain informative even when pointwise trajectory agreement degrades.

What carries the argument

The central object is the companion matrix C(α) whose last column is the positive-allocation coefficient vector α—the solution of the constrained least-squares problem that represents the final training snapshot as a convex combination of earlier snapshots. The shift structure of C means the model trajectory exactly matches the training data for the first n steps, after which it evolves by the linear recurrence encoded in α. The nonnegativity and normalization constraints are the load-bearing machinery: they turn the characteristic polynomial into one with a guaranteed spectral radius of one, which in turn makes the modal and finite-difference identities possible.

What would settle it

Apply the method to the linear growth map x_{k+1}=2x_k over a training window of n points. Because the positive-allocation constraints force the companion spectrum into the unit disk, the model trajectory will not grow, while the true trajectory doubles each step; the pointwise error will diverge, showing that the spectral guarantee does not imply predictive accuracy. More directly, on any validation trajectory, compute the one-step residual δ_m = ||F(ŷ_m) − ŷ_{m+1}||; if this residual is not small, the error decomposition (22) leaves the prediction error uncontrolled.

Watch

Extended reading notes

Core claim

The paper's central claim is Proposition 1: for a polynomial p(z) = z^n − Σ_{j=0}^{n−1} a_j z^j with nonnegative coefficients summing to one, every root lies in the closed unit disk and z=1 is a root. Therefore the companion matrix C built from the positive-allocation coefficients has spectral radius exactly one. The model trajectory ŷ_k = Y_train C^k e_1 reproduces the training snapshots for the first n steps and then follows a constant-coefficient recurrence that cannot grow unboundedly. When C is diagonalizable, the model trajectory has a modal form ŷ_k = Σ b_ℓ λ_ℓ^k ψ_ℓ, and the forward first and second differences become Σ b_ℓ(λ_ℓ−1)λ_ℓ^k ψ_ℓ and Σ b_ℓ(λ_ℓ−1)^2 λ_ℓ^k ψ_ℓ, respectively.

Load-bearing premise

The load-bearing premise is that recurrence coefficients obtained by fitting only the most recent snapshot as a convex combination of earlier snapshots will generate meaningful future states when applied as a constant-coefficient recurrence; the paper proves internal spectral properties but gives no control over the one-step residuals δ_m in Proposition 4, so pointwise accuracy is not guaranteed.

Editorial extensions

If this is right

  • Bounded model trajectories: any positive-allocation companion predictor has spectral radius 1, so the generated trajectory cannot blow up exponentially regardless of the underlying dynamics, distinguishing it from unconstrained linear fits.
  • Spectral filtering: finite differences of the model trajectory act as high-pass filters: modes with eigenvalues near 1 are suppressed in the first and second differences, so derivative-like diagnostics emphasize oscillatory or changing components.
  • Non-modal diagnostics: the C-based bounds can be evaluated directly from training data and C without eigendecomposition, so they remain computable even when C is defective or ill-conditioned.
  • Interpretability through sparsity: positive-allocation solutions tend to be sparse, allowing the recurrence to be read as 'the next snapshot is a weighted average of a few earlier snapshots'—a human-readable prediction mechanism.
  • Diagnostic value beyond accuracy: even when pointwise trajectory agreement degrades, the modal and finite-difference structure of the constructed recurrence may still track qualitative features such as growth, decay, and curvature.

Reading between the lines

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

  • The spectral containment result suggests a new regularization principle for data-driven linear predictors: constraining companion coefficients to the probability simplex is a parameter-free way to guarantee boundedness, which could be combined with other objectives to shape prediction horizons.
  • The finite-difference filter interpretation implies the method could serve as a data-driven differentiator or change-point detector: the second-difference diagnostic might flag regime changes in a signal even without an accurate forward model—an extension not explored in the paper.
  • A testable extension: for recurrent dynamics, lengthening the training window may concentrate positive-allocation coefficients on snapshots separated by the dominant period, yielding a data-driven period estimate from the sparse support of α.
  • The method's reliance on the convex-hull assumption suggests it will work best on recurrent or slowly varying trajectories; for strongly expanding or chaotic systems, the allocation residual will be large and the diagnostics will reflect the construction rather than the true dynamics.
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 / 5 minor

Summary. The paper introduces a 'positive-allocation companion predictor' for nonlinear dynamics. Given a training window of observable snapshots y_0,...,y_n, the method solves a constrained least-squares problem that represents the final snapshot y_n as a nonnegative, unit-sum combination of earlier snapshots. The resulting coefficients define a companion matrix C whose spectral properties are then used to generate a model trajectory and to derive modal and non-modal finite-difference diagnostics. The main analytical results are Proposition 1 (normalized nonnegative coefficients place the companion spectrum in the closed unit disk), the modal representation of the predictor and finite differences (Propositions A, 2, 3), C-based finite-difference bounds that avoid diagonalization (Proposition 5), and a standard discrete error decomposition (Proposition 4). Numerical experiments on the FitzHugh–Nagumo and SIR systems are presented as illustrations.

Significance. The mathematical core of the paper is correct: Proposition 1 is a simple and valid bound, and the modal and C-based finite-difference identities follow directly from the definitions. The idea of imposing nonnegativity and normalization on the recurrence coefficients at the construction stage is a neat way to obtain a priori spectral containment, and the C-based bounds (Proposition 5) are genuinely non-modal and computable from data without eigendecomposition. However, the significance of the work as a prediction method is not established. The paper explicitly disclaims accuracy guarantees, and the numerical section provides only qualitative comparisons with no quantitative error metrics, no baselines, and no convergence or robustness analysis. The spectral results are interesting as internal properties of the constructed model, but the paper does not demonstrate that these properties translate into predictive skill for nonlinear dynamics beyond the illustrative examples.

major comments (3)
  1. [Section 4 (Figs. 2–7)] The numerical evaluation is entirely qualitative. No prediction errors (e.g., RMSE, normalized error, or horizon-until-failure) are reported, and no comparison is made with DMD, EDMD, or other standard Koopman predictors. Statements such as 'captures several qualitative features' and 'pointwise agreement degrades' are not quantified. To support the claim that the method is a 'predictor' with practical value, the paper needs quantitative error metrics and at least one baseline comparison.
  2. [Algorithm 1 and Section 4] Several implementation parameters are left unspecified, including the Gaussian kernel bandwidth σ, the regularization parameter η, the coordinate scaling/centering procedure, the training-window length, and the quadratic-program solver. These choices can materially affect the allocation coefficients and the companion spectrum (Remark 4 acknowledges sensitivity). Without these details, the numerical experiments are not reproducible, and it is impossible to assess the robustness of the reported behavior.
  3. [Proposition 4 (Eq. (22))] The error decomposition in Eq. (22) separates the prediction error into an initial error term and accumulated one-step residuals δ_m, but it provides no control on δ_m. The allocation objective only constrains the first generated point ŷ_n ≈ y_n; it gives no bound on ŷ_{n+1}, ŷ_{n+2}, ... . Thus the central transfer assumption—that a convex fit to the last training snapshot yields a recurrence that tracks future dynamics—is unproven. The paper should either derive a bound on δ_m (e.g., in terms of the allocation residual, Lipschitz constant of F, or spectral properties of C) or explicitly reposition the method as a diagnostic tool rather than a predictor.
minor comments (5)
  1. [Section 4.1] Typo: 'e.g.,≈3.3×10 12 ompared with' should read 'compared'.
  2. [Section 4.1.1 / Remark 5] The paper defines forward second differences in Section 3 but the implementation uses a centered second difference. Remark 5 explains the index shift, but it would be cleaner to define the centered version in Section 3 to avoid confusion.
  3. [Section 2.2] Proposition A is not numbered in the same sequence as the other propositions; consider renumbering it or labeling it consistently.
  4. [Remark 4] The claim that 'decreasing the time step or extending the training window can increase the condition number' is given without quantitative support. A small table or figure showing condition numbers for the reported settings would be helpful.
  5. [Throughout] The paper repeatedly states that spectral containment 'does not by itself guarantee accurate prediction,' but the abstract and introduction still frame the construction as a predictor. The language could be tightened to distinguish the diagnostic contribution from predictive claims.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the spectral theorem and finite-difference identities are proved directly from coefficient constraints and matrix algebra; the unproven predictive-transfer step is openly labeled a limitation, not disguised as a derivation.

full rationale

The derivation chain is self-contained and non-circular. The allocation vector alpha is obtained from a constrained least-squares fit to the observed snapshot y_n; the companion identity C^k e1 = e_{k+1} for k<n means the model trajectory reproduces y_0,...,y_{n-1}, and C^n e1 = alpha makes yhat_n = Y_train alpha exactly the fitted convex combination. The paper does not present yhat_n as an independent prediction; it says only that 'When the allocation residual is small, by_256 provides an approximation of the observed snapshot y_256,' and repeatedly warns that spectral containment 'does not by itself guarantee accurate prediction.' Proposition 1 is a standalone elementary proof from a_j >= 0 and sum a_j = 1, not an input to the fitting; the eigenvalue 1 follows algebraically from p(1) = 1 - sum a_j. Propositions A, 2, 3, and 5 are algebraic consequences of C = Gamma Lambda Gamma^{-1} and of C^k(C-I)^r; they are not fitted to data. Proposition 4 is a standard Lipschitz error decomposition and is explicitly not a bound on the one-step residuals delta_m; that is a stated limitation of predictive accuracy, not a circular justification. The only overlapping-author citation, ref. [32], is used only to note observed sparsity in portfolio positive-allocation problems; it is not load-bearing for any theorem or prediction. No uniqueness theorem, ansatz, or fitted parameter is imported from the authors' prior work. Therefore no circular step can be exhibited.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central theorem needs only the simplex constraint on α; the numerical examples additionally rely on hand-chosen kernel bandwidth, regularization, scaling, observables, and training-window length, none of which are given precise values. No new physical or mathematical entities are postulated beyond the companion-predictor construction itself.

free parameters (6)
  • Allocation coefficients α* = not reported (n-dimensional QP solution)
    Output of the constrained least-squares problem; determines the recurrence and all numerical predictions. Not a free parameter of the theorem, but essential to the numerical examples.
  • Gaussian kernel bandwidth σ
    Required for the kernelized QP used in the numerical experiments; value not stated in the paper.
  • Regularization parameter η
    Added to the Gram matrix in Algorithm 1; value not reported.
  • Coordinate scaling/centering for FitzHugh–Nagumo
    Authors state data were centered and scaled per coordinate and that this changed the condition number from 1.8×10^4 to ≈3.3×10^12; the exact transform is not specified.
  • SIR observable augmentation (I²)
    Choice g(S,I,R)=[S,I,R,I^2] is ad hoc; no selection criterion or sensitivity analysis is provided.
  • Training window length = FHN: 51 snapshots (n=50); SIR: not stated
    Authors note training-window length affects conditioning; exact SIR window/sample count is not specified.
assumptions (5)
  • standard math The roots of the companion matrix C equal the roots of its characteristic polynomial p(z).
    Used implicitly in Section 2.1 and Remark 4 to identify eigenvalues of the companion matrix with the roots of the allocation polynomial.
  • standard math Submultiplicativity of induced matrix norms: ||AB|| ≤ ||A|| ||B||.
    Used in Propositions 3 and 5 to derive growth and finite-difference bounds.
  • domain assumption C is diagonalizable in the modal analysis.
    Propositions A, 2, and 3 assume diagonalizability; this is not guaranteed for companion matrices and fails when eigenvalues coincide (defective case).
  • domain assumption The true map F is Lipschitz with constant L and a fixed reconstruction from observables to states exists.
    Proposition 4's error decomposition requires this; no Lipschitz constant is estimated for the FHN or SIR examples.
  • domain assumption The kernel k is positive definite, making the quadratic program (13) well-posed.
    Assumed in the kernelized formulation (Section 2.1); the Gaussian kernel is used in experiments without explicit verification for the given data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Positive-Allocation Companion Predictors for Nonlinear Dynamics and Their Finite-Difference Diagnostics." pith.science (2026). https://pith.science/paper/KWPKQETA

@misc{pith2026260716529,
  author       = {Pith},
  title        = {Pith review of: Positive-Allocation Companion Predictors for Nonlinear Dynamics and Their Finite-Difference Diagnostics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KWPKQETA}},
  note         = {Machine review of arXiv:2607.16529}
}
abstract

We introduce a positive-allocation companion construction for Koopman-inspired finite-dimensional prediction of nonlinear dynamical systems. The method determines recurrence coefficients by representing a target observable snapshot as a nonnegative, normalized combination of earlier training snapshots. These coefficients define a companion matrix whose spectral structure is induced by the allocation constraints at the construction stage. We prove that normalized positive allocation places the companion spectrum in the closed unit disk and, because the coefficients sum to one, includes $1$ as an eigenvalue. Additionally, we develop modal and non-modal diagnostics for the resulting model trajectory. When the companion matrix is diagonalizable, the modal representation shows that first and second finite differences act as spectral filters through factors of $\lambda_\ell-1$ and $(\lambda_\ell-1)^2$. We also derive $C$-based finite-difference bounds that avoid diagonalization and can be evaluated directly from the training data and companion matrix. Numerical experiments on the FitzHugh--Nagumo and susceptible--infectious--recovered (SIR) models illustrate the behavior of the construction in oscillatory and transient dissipative settings. The examples demonstrate both the interpretability of the companion recurrence and its limitations, particularly when pointwise trajectory agreement degrades while finite-difference and modal diagnostics remain informative.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 2 canonical work pages

  1. [1]

    Proceedings of the National Academy of Sciences113(15), 3932–3937 (2016) https://doi.org/10.1073/pnas.1517384113 https://www.pnas.org/doi/pdf/10.1073/pnas.1517384113

    Brunton, S.L., Proctor, J.L., Kutz, J.N.: Discovering governing equations from data by sparse identification of nonlinear dynamical systems. Proceedings of the National Academy of Sciences113(15), 3932–3937 (2016) https://doi.org/10.1073/pnas.1517384113 https://www.pnas.org/doi/pdf/10.1073/pnas.1517384113

    doi:10.1073/pnas.1517384113 2016

  2. [2]

    Proceedings of the National Academy of Sciences116(45), 22445–22451 (2019) https://doi.org/10.1073/pnas.1906995116 https://www.pnas.org/doi/pdf/10.1073/pnas.1906995116

    Champion, K., Lusch, B., Kutz, J.N., Brunton, S.L.: Data-driven discovery of coordinates and governing equations. Proceedings of the National Academy of Sciences116(45), 22445–22451 (2019) https://doi.org/10.1073/pnas.1906995116 https://www.pnas.org/doi/pdf/10.1073/pnas.1906995116

    doi:10.1073/pnas.1906995116 2019

  3. [3]

    Nature Communications9(1), 4950 (2018) https://doi.org/10.1038/s41467-018-07210-0

    Lusch, B., Kutz, J.N., Brunton, S.L.: Deep learning for universal linear embeddings of nonlinear dynamics. Nature Communications9(1), 4950 (2018) https://doi.org/10.1038/s41467-018-07210-0

    doi:10.1038/s41467-018-07210-0 2018

  4. [4]

    In: Advances in Neural Information Processing Systems (NeurIPS), vol

    Chen, R.T.Q., Rubanova, Y., Bettencourt, J., Duvenaud, D.: Neural ordinary differential equations. In: Advances in Neural Information Processing Systems (NeurIPS), vol. 31 (2018). arXiv:1806.07366. https://arxiv.org/abs/1806.07366

    arXiv 2018

  5. [5]

    Pathak, J., Hunt, B., Girvan, M., Lu, Z., Ott, E.: Model-free prediction of large spatiotemporally chaotic systems from data: A reservoir computing approach. Phys. Rev. Lett.120, 024102 (2018) https://doi.org/10.1103/PhysRevLett.120.024102

    doi:10.1103/physrevlett.120.024102 2018

  6. [6]

    Prentice Hall PTR, Upper Saddle River, NJ (1999)

    Ljung, L.: System Identification: Theory for the User, 2nd edn. Prentice Hall PTR, Upper Saddle River, NJ (1999)

    1999

  7. [7]

    Nature Machine Intelligence1(5), 206–215 (2019) https://doi.org/10

    Rudin, C.: Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nature Machine Intelligence1(5), 206–215 (2019) https://doi.org/10. 1038/s42256-019-0048-x

    2019

  8. [8]

    Theoretical and Computational Fluid Dynamics31, 349–368 (2017) https://doi.org/10.1007/s00162-017-0432-2

    Hemati, M.S., Rowley, C.W., Deem, E.A., Cattafesta, L.N.: De-biasing the dynamic mode decom- position for applied Koopman spectral analysis of noisy datasets. Theoretical and Computational Fluid Dynamics31, 349–368 (2017) https://doi.org/10.1007/s00162-017-0432-2

    doi:10.1007/s00162-017-0432-2 2017

Show all 36 references
  1. [9]

    Flow, Turbulence and Combustion (2024) https://doi.org/10.1007/s10494-024-00595-7

    Mandler, H., Weigand, B.: Generalization limits of data-driven turbulence models. Flow, Turbulence and Combustion (2024) https://doi.org/10.1007/s10494-024-00595-7 . Published online November 9, 2024

    2024 doi

  2. [10]

    Journal of Nonlinear Science33(1), 14 (2023) https: //doi.org/10.1007/s00332-022-09862-1

    N¨ uske, F., Peitz, S., Philipp, F., Schaller, M., Worthmann, K.: Finite-Data Error Bounds for 25 Positive-Allocation Companion Predictions Flores and Pascoe Koopman-Based Prediction and Control. Journal of Nonlinear Science33(1), 14 (2023) https: //doi.org/10.1007/s00332-022-09862-1

    2023 doi

  3. [11]

    SIAM Review64(2), 229–340 (2022) https://doi.org/10.1137/21M1401243

    Brunton, S.L., Budiˇ si´ c, M., Kaiser, E., Kutz, J.N.: Modern Koopman theory for dynamical systems. SIAM Review64(2), 229–340 (2022) https://doi.org/10.1137/21M1401243

    2022 doi

  4. [12]

    Journal of Fluid Mechanics641, 115–127 (2009) https://doi.org/10.1017/S0022112009992059

    Rowley, C.W., Mezi´ c, I., Bagheri, S., Schlatter, P., Henningson, D.S.: Spectral analysis of nonlinear flows. Journal of Fluid Mechanics641, 115–127 (2009) https://doi.org/10.1017/S0022112009992059

    2009 doi

  5. [13]

    Annual Review of Fluid Mechanics45, 357–378 (2013) https://doi.org/10.1146/annurev-fluid-011212-140652

    Mezi´ c, I.: Analysis of fluid flows via spectral properties of the Koopman operator. Annual Review of Fluid Mechanics45, 357–378 (2013) https://doi.org/10.1146/annurev-fluid-011212-140652

    2013 doi

  6. [14]

    SIAM Journal on Applied Dynamical Systems24(2), 1150–1190 (2025) https://doi.org/10.1137/24M1662370

    Colbrook, M.J., Drysdale, C., Horning, A.: Rigged dynamic mode decomposition: Data-driven gen- eralized eigenfunction decompositions for Koopman operators. SIAM Journal on Applied Dynamical Systems24(2), 1150–1190 (2025) https://doi.org/10.1137/24M1662370

    2025 doi

  7. [15]

    In: Koyejo, S., Mohamed, S., Agarwal, A., Belgrave, D., Cho, K., Oh, A

    Kostic, V., Novelli, P., Maurer, A., Ciliberto, C., Rosasco, L., Pontil, M.: Learning dynamical systems via Koopman operator regression in reproducing kernel hilbert spaces. In: Koyejo, S., Mohamed, S., Agarwal, A., Belgrave, D., Cho, K., Oh, A. (eds.) Advances in Neural Infor...

    2022

  8. [16]

    Annual Reviews in Control52, 197–212 (2021) https://doi.org/10.1016/j.arcontrol.2021.09

    Bevanda, P., Sosnowski, S., Hirche, S.: Koopman operator dynamical models: Learning, analysis and control. Annual Reviews in Control52, 197–212 (2021) https://doi.org/10.1016/j.arcontrol.2021.09. 002

    2021 doi

  9. [17]

    Journal of Computational Dynamics1(2), 391–421 (2014) https://doi.org/10.3934/jcd.2014.1.391

    Tu, J.H., Rowley, C.W., Luchtenburg, D.M., Brunton, S.L., Kutz, J.N.: On dynamic mode decom- position: Theory and applications. Journal of Computational Dynamics1(2), 391–421 (2014) https://doi.org/10.3934/jcd.2014.1.391

    2014 doi

  10. [18]

    Journal of Nonlinear Science25(6), 1307–1346 (2015) https://doi.org/10.1007/s00332-015-9258-5

    Williams, M.O., Kevrekidis, I.G., Rowley, C.W.: A data–driven approximation of the Koopman operator: Extending dynamic mode decomposition. Journal of Nonlinear Science25(6), 1307–1346 (2015) https://doi.org/10.1007/s00332-015-9258-5

    2015 doi

  11. [19]

    SIAM Journal on Applied Dynamical Systems16(4), 2096– 2126 (2017) https://doi.org/10.1137/17M1125236

    Arbabi, H., Mezi´ c, I.: Ergodic theory, dynamic mode decomposition, and computation of spectral properties of the Koopman operator. SIAM Journal on Applied Dynamical Systems16(4), 2096– 2126 (2017) https://doi.org/10.1137/17M1125236

    doi

  12. [20]

    Journal of Computational Dynamics2(2), 247–265 (2015) https://doi.org/10

    Williams, M.O., Rowley, C.W., Kevrekidis, I.G.: A kernel-based method for data-driven Koopman spectral analysis. Journal of Computational Dynamics2(2), 247–265 (2015) https://doi.org/10. 3934/jcd.2015005

    2015

  13. [21]

    IEEE Control Systems Letters7(7) (2023) https://doi.org/10.1109/lcsys

    Morrison, Z., Russo, B.P., Lian, Y., Kamalapurkar, R.: Fault detection via occupation kernel prin- cipal component analysis. IEEE Control Systems Letters7(7) (2023) https://doi.org/10.1109/lcsys. 2023.3287568

    2023

  14. [22]

    arXiv preprint arXiv:2501.17976 (2025)

    Ait Yahia, I., Berrada, I.: KoopAGRU: A Koopman-based anomaly detection in time-series using gated recurrent units. arXiv preprint arXiv:2501.17976 (2025). Submitted Jan 2025

    2025 arXiv

  15. [23]

    Chaos: An Interdisciplinary Journal of Nonlinear Science22(4), 047510 (2012) https://doi.org/10.1063/1.4772195 26 Positive-Allocation Companion Predictions Flores and Pascoe

    Budiˇ si´ c, M., Mohr, R., Mezi´ c, I.: Applied Koopmanism. Chaos: An Interdisciplinary Journal of Nonlinear Science22(4), 047510 (2012) https://doi.org/10.1063/1.4772195 26 Positive-Allocation Companion Predictions Flores and Pascoe

    2012 doi

  16. [24]

    IEEE Transactions on Automatic Control65(12), 5114–5129 (2020) https://doi.org/10.1109/TAC

    Korda, M., Mezi´ c, I.: Optimal construction of Koopman eigenfunctions for prediction and control. IEEE Transactions on Automatic Control65(12), 5114–5129 (2020) https://doi.org/10.1109/TAC. 2020.2978039

    2020

  17. [25]

    Physica D: Nonlinear Phenomena406, 132416 (2020) https://doi.org/10.1016/j.physd.2020.132416

    Klus, S., N¨ uske, F., Peitz, S., Niemann, J.-H., Clementi, C., Sch¨ utte, C.: Data-driven approximation of the Koopman generator: Model reduction, system identification, and control. Physica D: Nonlinear Phenomena406, 132416 (2020) https://doi.org/10.1016/j.physd.2020.132416

    2020

  18. [26]

    (eds.): The Koopman Operator in Systems and Control: Concepts, Methodologies, and Applications

    Mauroy, A., Mezi´ c, I., Susuki, Y. (eds.): The Koopman Operator in Systems and Control: Concepts, Methodologies, and Applications. Lecture Notes in Control and Information Sciences, vol. 484. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-35713-9

    2020 doi

  19. [27]

    IEEE Transactions on Robotics (2021) https://doi.org/10.1109/ TRO.2021.3076581

    Mamakoukas, G., Castano, M., Tan, X., Murphey, T.: Derivative-based Koopman operators for real- time control of robotic systems. IEEE Transactions on Robotics (2021) https://doi.org/10.1109/ TRO.2021.3076581

    2021

  20. [28]

    Communications on Pure and Applied Mathematics77(1), 221–283 (2024) https://doi.org/10.1002/cpa.22125

    Colbrook, M.J., Townsend, A.: Rigorous data-driven computation of spectral properties of koopman operators for dynamical systems. Communications on Pure and Applied Mathematics77(1), 221–283 (2024) https://doi.org/10.1002/cpa.22125

    2024 doi

  21. [29]

    Journal of Fluid Mechanics955, 21 (2023) https://doi.org/10.1017/jfm.2022.1032

    Colbrook, M.J., Ayton, L.J., Sz˝ oke, M.: Residual dynamic mode decomposition: robust and verified koopmanism. Journal of Fluid Mechanics955, 21 (2023) https://doi.org/10.1017/jfm.2022.1032

    2023

  22. [30]

    Nonlinear Dynamics41(1), 309–325 (2005) https://doi.org/10.1007/s11071-005-2824-x

    Mezi´ c, I.: Spectral properties of dynamical systems, model reduction and decompositions. Nonlinear Dynamics41(1), 309–325 (2005) https://doi.org/10.1007/s11071-005-2824-x

    2005 doi

  23. [31]

    Journal of Financial and Quantitative Analysis27(4), 513–537 (1992)

    Best, M.J., Grauer, R.R.: Positively weighted minimum-variance portfolios and the structure of asset expected returns. Journal of Financial and Quantitative Analysis27(4), 513–537 (1992)

    1992

  24. [32]

    arXiv preprint arXiv:2412.18201 (2024)

    Hutinet, G., Pascoe, J.: Indices of quadratic programs over reproducing kernel Hilbert spaces for fun and profit. arXiv preprint arXiv:2412.18201 (2024)

    2024 arXiv

  25. [33]

    Soci- ety for Industrial and Applied Mathematics, Philadelphia, PA (2016)

    Kutz, J.N., Brunton, S.L., Brunton, B.W., Proctor, J.L.: Dynamic Mode Decomposition. Soci- ety for Industrial and Applied Mathematics, Philadelphia, PA (2016). https://doi.org/10.1137/1. 9781611974508 .https://epubs.siam.org/doi/abs/10.1137/1.9781611974508

    2016 doi

  26. [34]

    Biophys- ical Journal1(6), 445–466 (1961) https://doi.org/10.1016/S0006-3495(61)86902-6

    FitzHugh, R.: Impulses and physiological states in theoretical models of nerve membrane. Biophys- ical Journal1(6), 445–466 (1961) https://doi.org/10.1016/S0006-3495(61)86902-6

    1961 doi

  27. [35]

    Proceedings of the IRE50(10), 2061–2070 (1962) https://doi.org/10.1109/JRPROC.1962.288235

    Nagumo, J., Arimoto, S., Yoshizawa, S.: An active pulse transmission line simulating nerve axon. Proceedings of the IRE50(10), 2061–2070 (1962) https://doi.org/10.1109/JRPROC.1962.288235

  28. [36]

    Pro- ceedings of the Royal Society of London

    Kermack, W.O., McKendrick, A.G.: A contribution to the mathematical theory of epidemics. Pro- ceedings of the Royal Society of London. Series A115(772), 700–721 (1927) https://doi.org/10. 1098/rspa.1927.0118 27

    1927

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.