Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Learning dissipation and instability fields from chaotic dynamics

T0 review · 4 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read This paper claims that local dissipation and instability fields of a chaotic map can be estimated directly from a transition matrix built from orbit data, without knowing the governing equations.

desk verdict A_i is a clean, useful data-driven estimator of inverse dissipation, but the paper's claimed lower bound on the local Jacobian via B_j is actually an upper bound — the central advertised claim needs correcting. read the letter →

arxiv 2502.03456 v1 pith:6LIKJNUV submitted 2025-02-05 nlin.CD

Ludovico T Giorgini , Andre N Souza , Domenico Lippolis , Predrag Cvitanović , Peter Schmid This is my paper · ORCID
classification nlin.CD MSC 37M2537D4537C3037E05
keywords Perron-FrobeniusoperatortransitionmatrixJacobianestimationdissipationfieldinstabilitychaoticmapsdata-drivendynamicsstate-spacepartition
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 aims to establish that the local stability landscape of a chaotic map can be extracted from a purely statistical object: the transition matrix that records how often orbits move between cells of a partition. If correct, local dissipation and instability fields are available for systems whose equations are unknown, because the matrix is estimated directly from trajectory data. The paper derives two estimators from the Perron-Frobenius transport formula: the row sum $A_i$ approximates the inverse dissipation $\sum_k |J(f_k^{-1}(C_i))|^{-1}$, and the column maximum $B_j$ approximates the inverse Jacobian $|J(C_j)|^{-1}$ from above. Tests on one-dimensional maps and on two-dimensional coupled maps show that $A_i$ agrees with theory robustly while $B_j$ is reliable mainly where the map is strongly expanding.

What carries the argument

The machinery is the finite transition matrix $P_{ij}$ that approximates the Perron-Frobenius operator on a uniform partition $\{X_i\}$ of state space, with $P_{ij}$ the probability that mass in cell $j$ lands in cell $i$ in one step. The load-bearing step is to replace the integrand in the Perron-Frobenius transport formula by its value at the cell centroid $C_i$, which turns each matrix entry into a combination of inverse Jacobians and preimage transition probabilities. From that the row sum $A_i$ and the column maximum $B_j$ are exact algebraic targets: $A_i$ aggregates the inverse Jacobians over all preimages of $C_i$, while $B_j$ isolates the single dominant preimage and yields the bound $B_j\le |J(C_j)|^{-1}$.

What would settle it

Estimate $A_i$ and $B_j$ for a one-dimensional map with a critical point (where $|J|$ vanishes) using a partition that places the critical point inside a single cell. The analytical inverse Jacobian diverges at that cell while every transition-matrix entry satisfies $P_{ij}\le 1$, so $B_j$ cannot reproduce the analytical value; the size of that discrepancy, and whether $A_i$ also fails there, settles how much of the central claim survives.

Watch

Extended reading notes

Core claim

The central claim is that the local Jacobian of a chaotic map is encoded in the transition matrix. For a partition of state space into $N$ equal control volumes with centroids $C_i$, the empirical transition matrix $P$ obeys $A_i=\sum_j P_{ij}\approx \sum_k |J(f_k^{-1}(C_i))|^{-1}$, so the row sums are the local inverse dissipation field, and $B_j=\max_i P_{ij}\lesssim |J(C_j)|^{-1}$, so the column maxima give an upper bound on the inverse Jacobian and therefore a lower bound on local stretching. Because $P$ can be counted directly from orbit data, both fields are available without knowing the map. Numerical tests show $A_i$ matches the analytical dissipation field accurately even with noise, while $B_j$ is accurate where the map is strongly expanding and degrades in weakly unstable regions.

Load-bearing premise

The derivation assumes the partition is fine enough and the Jacobian smooth enough that the Jacobian is nearly constant on each cell, and that the Jacobian does not vanish; if either fails, the estimated fields are biased.

Editorial extensions

If this is right

  • For any discrete-time system with enough orbit data, the transition matrix alone yields a map of where infinitesimal perturbations grow and where phase-space volume contracts, without any model equation.
  • The $A_i$ field is insensitive to moderate observational noise, so it can serve as a data-driven check of whether a proposed model respects local conservation or dissipation.
  • The $B_j$ field, though weaker, localizes the strongly unstable regions where chaos control or sensitivity enhancement would be most effective.
  • The same estimators apply to experimental time series, provided the state space is partitioned consistently, making Jacobian fields accessible when equations are unavailable.
  • Because $A_i$ aggregates all preimages, the method handles maps with multiple branches, where explicit inverse maps are unavailable or expensive to compute.

Reading between the lines

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

  • A natural extension, not tested here, is to apply the estimators to the time-one map of a continuous flow or to a Poincaré section, which would bring ODE systems and experimental data into the same framework.
  • The robustness of $A_i$ relative to $B_j$ suggests a practical two-step strategy: use $B_j$ to identify unstable regions, then use $A_i$ to quantify the dissipation there; this is a workflow the authors gesture toward but do not work out.
  • Because entries of $P$ are bounded by 1, cells where $B_j$ saturates at its upper bound flag near-critical points with vanishing Jacobian; the failure mode could itself be used as a detector of critical loci.
  • A finite-sample error analysis comparing the convergence of $A_i$ and $B_j$ as the orbit length and partition resolution increase would sharpen the method's practical limits; the paper's numerics do not provide such scaling laws.
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

4 major / 5 minor

Summary. The paper proposes a data-driven method to estimate local dissipation and instability fields of a discrete-time dynamical system from a finite transition matrix approximating the Perron-Frobenius operator. The central objects are A_i = sum_j P_ij, which under a fine-partition and smooth-Jacobian approximation equals sum_k 1/|J(f^{-1}_k(C_i))|, and B_j = max_i P_ij, which the authors claim bounds the local inverse Jacobian. Numerical tests on one-dimensional maps (Ulam, continued fraction, cusp, Chebyshev) and two-dimensional coupled Chebyshev maps, with and without additive noise, compare both quantities with analytical Jacobians. The A_i estimates match the analytical values well, while B_j tracks the inverse Jacobian only in strongly unstable regions and saturates or underestimates in weakly unstable and contracting regions.

Significance. If the central claim holds, the method is valuable because it recovers local Jacobian information from orbit data alone, without fitted parameters and without knowledge of the governing equations. The A_i part is a clean and well-supported contribution, with a transparent derivation and extensive numerical validation, including a noise study. However, the B_j claim as stated is flawed: the inequality derived in the paper has the opposite direction from the advertised 'lower bound on the local Jacobian'. This is a load-bearing issue that affects the abstract and the interpretation of the instability-field estimates. The numerical evidence shows that B_j still provides a qualitative indicator in strongly unstable regions, so the paper can likely be repaired by restating the bound direction and the saturation limitations, but the current text must be corrected before acceptance.

major comments (4)
  1. [Section 2, Eq. (9) and Abstract] The direction of the claimed bound is reversed. Equation (9) states B_j = max_i P_ij <= min{|J(C_j)|^{-1}, 1}, so B_j is a lower bound on the inverse Jacobian 1/|J(C_j)|, not an upper bound, and the derived quantity 1/B_j satisfies 1/B_j >= max{|J(C_j)|, 1}, which is an upper bound on |J|. This contradicts the abstract's claim that 'a lower bound for the local Jacobian J(x) may also be estimated from the entries of the transfer matrix' and the sentence after Eq. (9) calling B_j 'an upper bound on the derivative of the forward map'. The underestimation of B_j in weakly unstable regions reported in Sections 3.1.5 and 3.2 is a manifestation of this inequality, not merely a convergence or noise artifact. The text should correct the bound direction or explicitly reframe B_j as a lower bound on 1/|J| that is informative only when |J| is sufficiently larger than one.
  2. [Section 2, derivation of Eq. (9)] The intermediate step B_j = max_i sum_k 1/|J(C_j)| p^k_ij ≈ 1/|J(C_j)| max_i p^{k*}_{ij} <= 1/|J(C_j)| drops the sum over k and assumes that the maximizing target cell receives the full image of the source cell. This approximation is uncontrolled: for |J(C_j)| < 1, the forward image of a source cell is smaller than a target cell, so the transition probability saturates at 1 and B_j becomes independent of 1/|J|. Thus Eq. (9) is not an equality-based estimator but a coarse upper bound with a gap that can be large. The paper should state this limitation explicitly and, ideally, give a quantitative condition on the partition size and Jacobian magnitude under which the gap is small.
  3. [Section 2, Eq. (5)] The coefficients p^k_ij are not defined unambiguously. The equality sum_j p^k_ij q^n_j := integral over X_i of rho_n(f^{-1}_k(x)) dx defines p^k_ij relative to a particular density rho_n, which does not imply that p^k_ij is a density-independent stochastic matrix. The subsequent interpretation of p^k_ij as the probability that f^{-1}_k(C_i) lies in X_j is a different, map-only object. The derivation of Eqs. (8) and (9) requires the latter; please either define p^k_ij directly as a Ulam-type transition probability and state the approximation, or prove that the two definitions agree in the fine-partition and smooth-density limit.
  4. [Appendix, after Eq. (22)] The statement that when the Jacobian vanishes, 'the integral ... <= 1 now provides an upper bound for max_i P_ij' is uninformative as written, because the probability bound P_ij <= 1 already holds by definition. The more relevant consequence is that the approximation in Eq. (22) fails, and no quantitative Jacobian estimate can be extracted from that cell. The non-vanishing-Jacobian assumption is load-bearing for both A_i and B_j and should be stated in the main text before Eq. (5), not only in the appendix.
minor comments (5)
  1. [Section 2] The word 'discretetization' near the end of Section 2 is a typo; it should read 'discretization'.
  2. [Reference [24]] The reference title contains 'Qantum'; it should be 'Quantum'.
  3. [Section 3.2] The word 'coeffficients' is a typo; it should read 'coefficients'.
  4. [Sections 2 and 3.1.4] The symbol N is used both for the number of control volumes in Eq. (1) and for the degree of the Chebyshev polynomial in Eq. (16). This creates confusion in Section 3.1.4, where N=100 control volumes are used alongside Chebyshev orders N=2,3,4,5; please use distinct symbols or explicitly state the context.
  5. [Figures 1 and 2 captions] The analytical B_j curves are capped at unity before comparison; the captions should state that the comparison is against min{1/|J|, 1}, so that the reader is aware that saturation is part of the displayed quantity.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the transition-matrix estimators are derived from first principles and validated against independent analytical Jacobians.

full rationale

The paper's central derivation is self-contained. The Perron-Frobenius relation in Eq. (4) is standard and is reproduced in the text rather than merely imported; the coarse-graining approximation leading to Eq. (5) is stated explicitly, and the definitions of A_i and B_j in Eqs. (8) and (9) follow from the transition matrix by algebraic manipulation, not by fitting or by assuming the target Jacobian values. The numerical comparisons in Section 3 use independently computed analytical Jacobians (Eqs. (10)-(18) and the 2D coupled Chebyshev Jacobian), so the data-derived quantities are not forced to match by construction. No parameter is fitted to a subset of data and then renamed a prediction. The self-citations to ChaosBook (Refs. [24,43]) supply background formulas for the Perron-Frobenius operator and example maps, but the load-bearing equations are derived in the paper, and the cited material does not presuppose the paper's claimed result. The citation to Ref. [44] for the properties of coupled Chebyshev maps is used only to justify the choice of numerical example, not to establish the method. The known limitation that B_j gives a bound whose direction is misstated in places (the abstract calls it a lower bound on the Jacobian while Eq. (9) yields B_j ≤ min{|J|^{-1},1}) is a correctness or interpretation issue, not a circularity: the data-derived B_j is not constructed from the analytical 1/|J|. The paper is therefore self-contained against external analytical benchmarks, and no circular step can be exhibited.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard transfer operator results plus two domain assumptions: fine partition with smooth Jacobian, and non-vanishing Jacobians. No new physical entities are introduced. The numerical hyperparameters N, sigma, and a are chosen by hand but are not fitted to the comparison targets.

free parameters (3)
  • Number of control volumes N = 100 in 1D; 100x100 grid in 2D
    Partition resolution chosen by hand; the approximation in Eq. (5) and the quality of A_i and B_j depend on it, and no convergence study over N is reported.
  • Noise amplitude sigma = 0, 0.001, 0.002, 0.003 in 1D; 0 in 2D
    Added to test robustness; values chosen ad hoc and no systematic noise scaling is derived.
  • Coupling strength a = 0.01
    Weak-coupling regime chosen so the 2D map is locally expanding; no variation over a is reported.
assumptions (5)
  • standard math Perron-Frobenius operator evolution formula: rho_{n+1}(x) = sum_k rho_n(f^{-1}_k(x)) / |J(f^{-1}_k(x))|.
    Taken from ChaosBook refs [24,37,43] and used as the starting point for the discretization in Section 2.
  • domain assumption The partition is sufficiently fine and the Jacobian sufficiently smooth that the Jacobian can be approximated by its value at the centroid C_i in Eq. (5).
    Invoked explicitly in Section 2 before Eq. (5); if violated, both A_i and B_j are biased.
  • ad hoc to paper For the B_j bound, all preimages f^{-1}_k(C_i) that fall in X_j can be approximated by C_j.
    Used in the derivation of Eq. (9); this is the main source of B_j's underestimation in weakly unstable regions.
  • domain assumption Non-vanishing Jacobians |J(f^{-1}_k(C_i))| and |J(C_j)|.
    Appendix states all expressions rely on this; if the Jacobian vanishes, approximation (22) no longer holds and only an upper bound is obtained.
  • standard math Transition matrix column normalization sum_i P_ij = 1 and bounds 0 <= P_ij <= 1.
    Properties of the Ulam discretization, used in Eq. (9).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning dissipation and instability fields from chaotic dynamics." pith.science (2026). https://pith.science/paper/6LIKJNUV

@misc{pith2026250203456,
  author       = {Pith},
  title        = {Pith review of: Learning dissipation and instability fields from chaotic dynamics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6LIKJNUV}},
  note         = {Machine review of arXiv:2502.03456}
}
read the original abstract

To make predictions or design control, information on local sensitivity of initial conditions and state-space contraction is both central, and often instrumental. However, it is not always simple to reliably determine instability fields or local dissipation rates, due to computational challenges or ignorance of the governing equations. Here, we construct an alternative route towards that goal, by estimating the Jacobian of a discrete-time dynamical system locally from the entries of the transition matrix that approximates the Perron-Frobenius operator for a given state-space partition. Numerical tests on one- and two-dimensional chaotic maps show promising results.

Figures

Figures reproduced from arXiv: 2502.03456 by the authors.

Figure 1
Figure 1. Ulam, continued fraction and cusp maps: Plots of Ai and Bj (left and right column, respectively) derived from the transition matrix (depicted with red lines), along with their analytical estimates (shown in blue), for the Ulam, continued fraction, and cusp maps. Noise amplitudes of σ = 0, 0.001, 0.002, 0.003 have been applied to the maps. Although the inverse of the two-dimensional map does not ad￾mit a closed-form … view at source ↗
Figure 2
Figure 2. One-dimensional Chebyshev map: Plots of Ai and Bj derived from the transition matrix (red lines), along with their analytical estimates (blue lines), for the Chebyshev map defined in Eq. (16) with N = 2, 3, 4, and 5, and σ = 0. mation. This verification would be accomplished by using f −1 to estimate the dissipation A, followed by a matching of this result to A, this time as derived from the transition matrix via Eq… view at source ↗
Figure 3
Figure 3. Two-dimensional coupled Chebyshev map. Heat maps and scatter plots comparing the Ai values from the transition matrix with their analytical counterparts for the map in Eq. (19), using N = 2, 3, 4 and coupling parameter a = 0.01. Below the heatmaps, we compare Ai as a function of the x coordinate for different values of y (y = −0.98, −0.9, 0.8 for N = 2 and y = −0.98, −0.96, 0.8 for N = 3, 4), obtained through the tw… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Two-dimensional coupled Chebyshev map. Same as Fig. (3), but for Bj . The values of y used in the plots in the last row are y = −0.98, −0.5, −0.3 for N = 2, y = −0.98, −0.7, −0.4 for N = 3 and y = −0.4, −0.1, −0.04 for N = 4. (2024) A1. doi:10.1017/jfm.2024.658. [30] A…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Reduced-Order Modeling of Cyclo-Stationary Time Series Using Score-Based Generative Methods

    nlin.CD 2025-08 conditional novelty 4.0 of 10

    A score-based generative model with an annual-cycle phase coordinate reproduces cyclo-stationary climate statistics from 20 principal components of a PlaSim simulation.

Reference graph

Works this paper leans on

45 extracted references · 41 canonical work pages · cited by 1 Pith paper

  1. [1]

    Lejarza, M

    F. Lejarza, M. Baldea, Data-driven discovery of the governing equations of dynamical systems via moving horizon optimization, Scientific Re- ports 12 (1) (2022) 11836

    work page 2022

  2. [2]

    Holmes, J

    P. Holmes, J. L. Lumley, G. Berkooz, Turbulence, Coherent Structures, Dynamical Systems and Symmetry, Cambridge University Press, 2012

    work page 2012

  3. [3]

    S. L. Brunton, J. L. Proctor, J. N. Kutz, Discovering governing equations from data: Sparse identification of nonlinear dynamical systems, Proceed- ings of the National Academy of Sciences 113 (15) (2016) 3932–3937

    work page 2016

  4. [4]

    P. J. Schmid, Dynamic mode decomposition of numerical and experimen- tal data, Journal of Fluid Mechanics 656 (2010) 5–28

    work page 2010

  5. [5]

    Santos Guti ´errez, V

    M. Santos Guti ´errez, V . Lucarini, M. D. Chekroun, M. Ghil, Reduced- order models for coupled dynamical systems: Data-driven methods and the Koopman operator, Chaos: An Interdisciplinary Journal of Nonlinear Science 31 (5) (2021)

    work page 2021

  6. [6]

    A. J. Chorin, P. Stinis, Problem reduction, renormalization, and mem- ory, Communications in Applied Mathematics and Computational Sci- ence 1 (1) (2007) 1–27

    work page 2007

  7. [7]

    Moehlis, T

    J. Moehlis, T. R. Smith, P. Holmes, H. Faisst, Models for turbulent plane Couette flow using the proper orthogonal decomposition, Physics of Flu- ids 16 (6) (2004) 2011–2024

    work page 2004

  8. [8]

    Kantz, T

    H. Kantz, T. Schreiber, Nonlinear Time Series Analysis, Cambridge Uni- versity Press, 2004. 6 Figure 3: Two-dimensional coupled Chebyshev map. Heat maps and scatter plots comparing theAi values from the transition matrix with their analytical counterparts for the map in Eq. (19), using N = 2, 3, 4 and coupling parameter a = 0.01. Below the heatmaps, we co...

    work page 2004

Show all 45 references
  1. [9]

    L. T. Giorgini, S. H. Lim, W. Moon, J. S. Wettlaufer, Precursors to rare events in stochastic resonance, Europhysics letters 129 (4) (2020) 40003

    2020

  2. [10]

    S. H. Lim, L. T. Giorgini, W. Moon, J. S. Wettlaufer, Predicting critical transitions in multiscale dynamical systems using reservoir computing, Chaos: An Interdisciplinary Journal of Nonlinear Science 30 (123126) (2020)

    2020

  3. [11]

    E. N. Lorenz, Deterministic nonperiodic flow, Journal of the Atmospheric Sciences 20 (2) (1963) 130–141

    1963

  4. [12]

    Schneider, S

    T. Schneider, S. Lan, A. Stuart, J. Teixeira, Earth system modeling 2.0: A blueprint for models that learn from observations and targeted high- resolution simulations, Geophysical Research Letters 44 (24) (2017) 12– 396

    2017

  5. [13]

    L. T. Giorgini, W. Moon, N. Chen, J. S. Wettlaufer, Non-Gaussian stochastic dynamical model for the El Ni ˜no Southern Oscillation, Physi- cal Review Research 4 (2) (2022) L022065

    2022

  6. [14]

    L. T. Giorgini, S. H. Lim, W. Moon, N. Chen, J. S. Wettlaufer, Modeling the El Ni ˜no Southern Oscillation with neural di fferential equations, in: Proceedings of the 38th International Conference on Machine Learning, 2021, p. 19

    2021

  7. [15]

    T. N. Palmer, Predicting uncertainty in forecasts of weather and climate, Reports on Progress in Physics 63 (2) (2000) 71–116

    2000

  8. [16]

    A. J. Majda, J. Harlim, Filtering Complex Turbulent Systems, Cambridge University Press, 2012

    2012

  9. [17]

    Keyes, L

    N. Keyes, L. Giorgini, J. Wettlaufer, Stochastic paleoclimatology: Mod- eling the EPICA ice core climate records, Chaos: An Interdisciplinary Journal of Nonlinear Science 33 (9) (2023)

    2023

  10. [18]

    Baldovin, F

    M. Baldovin, F. Cecconi, A. Provenzale, A. Vulpiani, Extracting causa- tion from millennial-scale climate fluctuations in the last 800 kyr, Scien- tific Reports 12 (1) (2022) 15320

    2022

  11. [19]

    Vissio, V

    G. Vissio, V . Lucarini, A proof of concept for scale-adaptive parametriza- tions: The case of the Lorenz’96 model, Quarterly Journal of the Royal Meteorological Society 144 (710) (2018) 63–75

    2018

  12. [20]

    L. T. Giorgini, K. Deck, T. Bischo ff, A. Souza, Response theory via generative score modeling, Phys. Rev. Lett. 133 (2024) 267302. doi: 10.1103/PhysRevLett.133.267302

    2024 doi

  13. [21]

    Black, M

    F. Black, M. Scholes, The pricing of options and corporate liabilities, Journal of Political Economy 81 (3) (1973) 637–654

    1973

  14. [22]

    A. J. Lotka, Elements of physical biology, Science 61 (1576) (1925) 400– 405

    1925

  15. [23]

    S. H. Strogatz, Nonlinear Dynamics and Chaos: With Applications to Physics, Biology, Chemistry, and Engineering, Westview Press, 1994

    1994

  16. [24]

    Cvitanovi ´c, R

    P. Cvitanovi ´c, R. Artuso, R. Mainieri, G. Tanner, G. Vattay, N. Whelan, A. Wirzba, Chaos: Classical and Qantum, Niels Bohr Inst., Copenhagen, 2025. URL https://ChaosBook.org

    2025

  17. [25]

    Vulpiani, Chaos: From Simple Models to Complex Systems, World Scientific, 2010

    A. Vulpiani, Chaos: From Simple Models to Complex Systems, World Scientific, 2010

    2010

  18. [26]

    A. N. Souza, T. Lutz, G. R. Flierl, Statistical non-locality of dynamically coherent structures, Journal of Fluid Mechanics 966 (2023) A44

    2023

  19. [27]

    Geogdzhayev, A

    G. Geogdzhayev, A. N. Souza, R. Ferrari, The evolving butterfly: Statis- tics in a changing attractor, Physica D: Nonlinear Phenomena (2024) 134107

    2024

  20. [28]

    L. T. Giorgini, A. N. Souza, P. J. Schmid, Reduced Markovian models of dynamical systems, https://arXiv.org/abs/2308.10864 (2023)

    2023 arXiv

  21. [29]

    A. N. Souza, Representing turbulent statistics with partitions of state space. part 1. theory and methodology, Journal of Fluid Mechanics 997 7 Figure 4: Two-dimensional coupled Chebyshev map. Same as Fig. (3), but for B j. The values of y used in the plots in the last row are...

    2024 doi

  22. [30]

    A. N. Souza, Representing turbulent statistics with partitions of state space. Part 2. The compressible Euler equations, Journal of Fluid Me- chanics 997 (2024) A2. doi:10.1017/jfm.2024.657

    2024 doi

  23. [31]

    A. N. Souza, S. Silvestri, A modified bisecting K-means for approxi- mating transfer operators: Application to the Lorenz equations, https: //arXiv.org/abs/2412.03734 (2024)

    2024 arXiv

  24. [32]

    Budi ˇsi´c, R

    M. Budi ˇsi´c, R. Mohr, I. Mezi ´c, Applied Koopmanism, Chaos: An Inter- disciplinary Journal of Nonlinear Science 22 (2012) 047510

    2012

  25. [33]

    Zhang, H

    C. Zhang, H. Li, Y . Lan, Phase space partition with Koopman analy- sis, Chaos: An Interdisciplinary Journal of Nonlinear Science 32 (2022) 063132

    2022

  26. [34]

    M. Buhl, M. B. Kennel, Statistically relaxing to generating partitions for observed time-series data, Physical Review E 71 (2005) 046213

    2005

  27. [35]

    Asllani, J

    M. Asllani, J. D. Challenger, F. S. Pavone, L. Sacconi, D. Fanelli, The the- ory of pattern formation on directed networks, Nature Communications 5 (2014) 4517

    2014

  28. [36]

    M. L. Kringelbach, G. Deco, Brain states and transitions: Insights from computational neuroscience, Cell Reports 32 (2020) 108128

    2020

  29. [37]

    Cvitanovi ´c, R

    P. Cvitanovi ´c, R. Artuso, L. Rondoni, E. A. Spiegel, Transporting densi- ties, in: P. Cvitanovi´c, R. Artuso, R. Mainieri, G. Tanner, G. Vattay (Eds.), Chaos: Classical and Quantum, Niels Bohr Inst., Copenhagen, 2025, p. 383. URL https://ChaosBook.org/paper.shtml#measure

    2025

  30. [38]

    D. J. Driebe, Fully Chaotic Maps and Time Reversal Symmetry, Springer, 1999

    1999

  31. [39]

    E. Ott, C. Grebogi, J. A. Yorke, Controlling chaos, Physical Review Let- ters 64 (11) (1990) 1196–1199

    1990

  32. [40]

    A. Wolf, J. B. Swift, H. L. Swinney, J. A. Vastano, Determining Lyapunov exponents from a time series, Physica D: Nonlinear Phenomena 16 (3) (1985) 285–317

    1985

  33. [41]

    J. M. T. Thompson, H. B. Stewart, Nonlinear Dynamics and Chaos: Ge- ometrical Methods for Scientists and Engineers, John Wiley and Sons, 2002

    2002

  34. [42]

    Wiggins, Introduction to Applied Nonlinear Dynamical Systems and Chaos, Springer, 2003

    S. Wiggins, Introduction to Applied Nonlinear Dynamical Systems and Chaos, Springer, 2003

    2003

  35. [43]

    C. Beck, F. Schl ¨ogl, Y . L. Klimontovich, Thermodynamics of chaotic sys- tems: An introduction, Physics-Uspekhi 37 (7) (1994) 713

    1994

  36. [44]

    Dettmann, D

    C. Dettmann, D. Lippolis, Periodic orbit theory of two coupled Tchebyscheff maps, Chaos, Solitons & Fractals 23 (1) (2005) 43–54. doi:10.1016/j.chaos.2004.04.017

    2005 doi

  37. [45]

    J. Yan, C. Beck, Information shift dynamics described by Tsallis q = 3 entropy on a compact phase space, Entropy 24 (2022) 1671. doi:10.3 390/e24111671. Appendix: Alternative derivation of the expressions for Ai and B j Let us write a density in the state space as a sum of con...

    2022

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.