Pith. sign in

REVIEW 4 major objections 4 minor 125 references

fSRD: Fuzzy Spectral Region Decomposition -- Automated Multi Operator Koopman Representations via an Adaptive Spectral Learning Architecture

T0 review · 4 major / 4 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 This paper claims that a nonlinear or chaotic system that admits no global finite-dimensional Koopman representation can nonetheless be reconstructed accurately as a fuzzy union of locally fitted linear Koopman operators, chosen automatical

desk verdict Novel architecture, unsupported central claim: the fuzzy-tree local DMD idea is worth a serious look, but the topological transform breaks the time axis and the 'Koopman representation' conclusion doesn't follow. read the letter →

arxiv 2607.17990 v1 pith:2KZTRSI5 submitted 2026-07-20 cs.LG math.DS

Charles Bokor , Mark Cary , Denise Morrey , Fabrizio Bonatesta This is my paper · ORCID
classification cs.LGmath.DS
keywords KoopmanoperatorDynamicModeDecompositionfuzzylocalmodelnetworksinvariantnonlineardynamicschaoticsystemsspectrallearningrepresentation
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

fSRD is a fully automated method for estimating Koopman representations not as one global operator but as a collection of locally valid linear operators. The paper claims this works even for systems whose Koopman operator has no finite set of global eigenfunctions, such as chaotic systems and systems with multiple attractors. A fuzzy tree automatically partitions the data matrix into regions, fits a dynamic-mode-decomposition model inside each region, and recombines the results into a single reconstruction. If correct, this would make Koopman-style linear representation practical for many nonlinear and high-dimensional problems without hand-crafted observables or prior system knowledge.

What carries the argument

The core mechanism is 'invariant decomposition': representing a Koopman operator by a disjoint union of restricted spectral integrals, each corresponding to a locally invariant subspace. This is implemented by a fuzzy tree whose sigmoidal splits define validity functions for each region; a local topological transform stretches each non-rectangular region to a rectangular matrix so that a truncated singular-value decomposition and dynamic mode decomposition can be fit; and a BIC-based pruning step removes redundant partitions to keep the model parsimonious.

What would settle it

Take a synthetic system with two known invariant regions separated by an oblique boundary, compute the true local Koopman eigenvalues in each region, apply fSRD's topological transform and dynamic mode decomposition, and compare the recovered eigenvalues and reconstruction error; if the interpolated temporal axis changes the eigenvalues or fails to reconstruct the original boundary region, the central claim is refuted.

Watch

Extended reading notes

Core claim

The central claim is that a nonlinear system that lacks a global linear Koopman representation can still be represented exactly enough by a finite collection of local Koopman operators, one per dynamically invariant region. fSRD discovers these regions automatically through a fuzzy hierarchical tree, applies dynamic mode decomposition locally after stretching each non-rectangular region into a rectangle, and then maps the local spectral models back to the original coordinates. The paper demonstrates this on chaotic oscillators, a bifurcation scenario, and a high-dimensional solar imagery sequence, reporting high reconstruction accuracy with interpretable region placement and modal decomposit

Load-bearing premise

The claim stands or falls on whether stretching a non-rectangular region of data into a rectangle by interpolation, and fitting a linear operator there, preserves the spectral content of the original local dynamics.

Editorial extensions

If this is right

  • Systems with continuous Koopman spectra, which ordinarily resist finite-dimensional DMD, can be reconstructed with high accuracy by partitioning into local operators.
  • The method requires no prior knowledge about the system's attractors, basins, or observables, extending Koopman-based modeling to broader data-driven applications.
  • The fuzzy tree provides interpretable region boundaries and local modal decompositions, revealing regime shifts and invariant structure directly from data.
  • The same algorithm applies across data-rich and data-limited regimes, with noise robustness up to the point where system structure is obscured.

Reading between the lines

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

  • Inference: If the topological transform indeed preserves local Koopman spectra, fSRD's tree could serve as a generic nonlinear sequence model, potentially rivaling recurrent or transformer architectures for structured matrix data while offering interpretable linear dynamics.
  • Inference: The paper leaves open how to choose observables beyond the identity; enriching each local region with learned or hand-crafted observables could collapse continuous spectra further, but would trade off the automation that makes fSRD attractive.
  • Inference: A direct test of the spectral-preservation assumption would be to compare the eigenvalues recovered from a single known invariant region with and without the interpolation step; if they differ significantly, the reconstructive accuracy may come from interpolation rather than from true local Koopman structure.
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 / 4 minor

Summary. The paper introduces fSRD, a fuzzy hierarchical tree architecture that partitions a data matrix X into overlapping regions via sigmoidal activation functions, fits a local DMD spectral model to each region after applying a 'topological transform' ϑ, and aggregates the local models with fuzzy weights. The authors claim that this constitutes an automated multi-operator finite Koopman representation for nonlinear, chaotic, and high-dimensional systems, and they support the claim with case studies on the Duffing oscillator, the Lorenz system (including a bifurcation study), and solar imagery. The paper also conjectures universal-approximator capability for sequential data matrices, while explicitly noting that a proof is beyond the scope of the work (Section 2.9).

Significance. If the core assertion were established — that fuzzy matrix-coordinate splits correspond to restricted spectral integrals of the Koopman operator and that the topological transform ϑ preserves the local spectral content — the contribution would be significant: a fully automated, prior-free, interpretable alternative to global Koopman methods and black-box sequence models. The paper is commendably transparent about several limitations: Section 2.9 admits the universality claim is unproved, Section 3.2.2 documents difficulties with pure continuous spectra, and Section 3.3 notes that the regularization procedure is not yet scalable. The case studies are extensive and include a high-dimensional real-world dataset. However, the paper does not ship machine-checked proofs or code, and the main text defers many key derivations to supplementary material and to a thesis [15] that is 'submitted, availability pending.' The significance is therefore entirely conditional on resolving the spectral-preservation question.

major comments (4)
  1. [§2.6.3, Eq. (38)] The topological transform ϑ is the load-bearing step of the method. It stretches each row of the non-rectangular region Xi to the full width q before DMD is applied, and then maps the fitted model back by ϑ^{-1}. This is a per-row reparameterization of the time axis: rows with different active interval lengths receive different stretch factors, so the fitted DMD eigenvalues ω_i are defined in a row-dependent stretched coordinate, not in the original evolution time. No theorem or numerical diagnostic establishes that ϑ preserves the Koopman spectrum of the local dynamics. The inverse interpolation only guarantees that the reconstructed training rows match the input, which explains why in-sample NRMSE can be small. Without an argument that ϑ is a local conjugacy or at least a spectral-preserving isomorphism, the operators in Eq. (38) cannot be called finite Koopman representations of the o
  2. [§2.2, Eqs. (30)–(36)] The paper asserts a correspondence between fuzzy splits in data-matrix coordinates and restricted spectral integrals 1_{[a,b]} dE(ω)g of the Koopman operator. The indicator 1_{[a,b]} is a frequency-domain multiplier, while the sigmoidal membership functions ∅_i(U) act on row and column indices of X. No mapping between these two domains is provided, and Eq. (36) is presented as an equality/approximation without proof. This is not a purely formal gap: the split positions are optimized by PSO to minimize reconstruction error, so the resulting partitions may simply carve the training matrix into blocks that are easy to fit linearly. The claimed 'invariant decomposition' and the partition categories I, cS, G, pO in Eqs. (36) and (48) are therefore not established.
  3. [§3, all case studies] All reported quantitative results (e.g., Duffing NRMSE 0.026, Lorenz Hankel 0.0068, Lorenz bifurcation 0.0377, solar 0.03) are reconstructions of the same data matrix X that was used to fit the local DMD models and to optimize the fuzzy split positions via PSO. This is an in-sample fit. A flexible piecewise-linear interpolator would also achieve small reconstruction error on training data. The abstract and Section 4 use terms such as 'strong predictive accuracy' and 'one shot model robustness,' but no held-out forecasting, cross-validation, or train/test split is reported anywhere in the manuscript. Without out-of-sample evaluation, the empirical results do not distinguish a learned Koopman representation from a high-capacity interpolator.
  4. [§2.9 and §4] The paper explicitly states that a proof of universal approximator capability is beyond its scope, which is acceptable for a conjecture. However, Sections 2.9 and 4 repeatedly assert that the case studies 'suggest' universality and that fSRD is 'a general alternative to a broad class of non-parametric estimators.' Given the absence of any generalization test and the unproven spectral-preservation step, these statements overstate the evidence. The authors should either supply a proof under explicit assumptions, or clearly restrict the claims to in-sample reconstruction by a heuristic piecewise-DMD architecture.
minor comments (4)
  1. [Eq. (14)] The last column of Y2 is written as xt_{m-1}; this appears to be a typo and should presumably be y_{t_{m-1}}.
  2. [§3.2.2] The text refers to 'figure-6.18-(a)' and 'figure-6.33-(a)', which appear to be thesis numbering, not the numbering of this manuscript. Please update all such cross-references.
  3. [Eq. (51)] The notation B^{-1} - δ B^{-2}, and the statement that the superscript for symmetric matrices refers to multiplication, is confusing and should be defined more carefully; the domain of the matrix B is not stated before the equation.
  4. [Throughout] Many key derivations (e.g., SM-2, SM-6, SM-8, SM-11, SM-12, SM-14, SM-15) are relegated to supplementary material, and reference [15] is a thesis listed as 'submitted, availability pending.' For a self-contained review, the main text should include at least the essential steps supporting Eqs. (41) and (51), as these are used to justify the pruning and regularization procedures.

Circularity Check

1 steps flagged · score 6.0 of 10

Reported 'predictive accuracy' is in-sample NRMSE minimized by the same split/DMD fitting procedure; otherwise the Koopman architecture has independent content.

  1. fitted input called prediction [Abstract; §2.3 Eq. (37); §2.5.1 Eq. (41); §2.5.2; §3.1.2 (Fig. 11)]
    "Empirical results across canonical chaotic systems ... demonstrate strong predictive accuracy ... X≈ eX= pX i=1 ∅i(U)⊙ efK,i(Xi) (37) ... N RM SEi = ∥ϑ(∅i(U)⊙X In,i )−Φ idiag(bi)T(ωDM D,i)∥F ∥ϑ(∅i(U)⊙X In,i )−µJ f,q ∥F (41) ... Figure-11 illustrates the application of fSRD on the identical set of Monte Carlo sampling (figure-10), improving performance withN RM SE∼0.026, i.e., 97.4% accuracy."

    The global model (37) is a weighted sum of local DMD reconstructions, and the NRMSE in (41) is both the split-selection/optimization objective (via (42) and PSO in §2.5.2) and the headline accuracy reported in §3 (e.g., 0.026). Since the same X is used to fit the DMD operators and to compute the reported errors, the reported 'highly accurate reconstructions'/'strong predictive accuracy' measure training fit, not a prediction or an independent Koopman property. No held-out forecasting test is presented, and the paper's own §2.9 concedes universality is unproved. Hence the empirical support for the central claim reduces by construction to the fit objective.

full rationale

The paper's methodological chain (Koopman spectral restriction, local DMD, fuzzy aggregation) is not itself circular: it builds on external DMD/Koopman results and standard LMN machinery, and the text explicitly labels the universal-approximation claim as a hypothesis ('A proof of this universal approximator capability was beyond the scope of this work'). The self-citations to [15], [22], and [23] are used for implementation details and derivation provenance rather than as an imported theorem that forces the main conclusion. However, the headline empirical evidence is circular in the narrower sense: the reported NRMSEs are computed on the same data matrix used to fit the local DMD operators and to optimize the fuzzy split positions via the same NRMSE criterion. The abstract's 'strong predictive accuracy' is thus supported only by in-sample reconstructions that are forced by the fitting procedure. Because the architecture has independent representational content beyond this fit, the score is 6 rather than 8–10; because the central empirical demonstration reduces to a training objective, it is not 0–2.

Assumptions & free parameters 7 free parameters · 6 assumptions · 1 invented entities

The central method depends on several fitted quantities (splits, SVD truncation, ridge parameters) and on the unproven assumption that fuzzy coordinate regions and the topological transform preserve Koopman invariance. No new physical entities are introduced; the invented 'invariant decomposition' is a conceptual construct without independent evidence.

free parameters (7)
  • split coefficient vectors v_j = optimized via PSO per split
    Position and direction of each sigmoid split are fit to minimize the weighted NRMSE cost (Eqs. 41–42).
  • sigmoid slope τ_j = not specified in main text
    Appears in Eq. (45) and controls the sharpness of fuzzy boundaries; details deferred to SM-22.
  • activation cutoff η = not specified
    Used to define non-zero support of membership functions and hence local region extraction (Section 2.6.1); chosen for numerical stability.
  • SVD truncation threshold r = 1e-4 in all experiments
    Hand-chosen rank threshold used in exact-DMD local models to handle rank deficiency.
  • ridge parameter δ_BIC = data-driven fixed-point per region
    Computed per region from Eq. (51) via fixed-point iteration; controls SVD regularization and local dimensionality.
  • BIC complexity scaling Θ = Θ=3 in factorial experiment
    Manual scaling parameter in the adapted BIC pruning formula (SM-8), which changes the number of retained regions.
  • PSO hyperparameters = unknown (SM-22)
    Particle swarm settings affect final split locations and are not given in the reviewed text.
assumptions (6)
  • domain assumption A Koopman operator exists for the observed system and can be approximated by finite-dimensional DMD on local regions.
    Standard Koopman/DMD theory is invoked in Sections 1.2–1.5 and used as the local model inside every fSRD region.
  • ad hoc to paper Fuzzy splits in the row/column indices of X correspond to invariant interfaces and restricted spectral integrals of the Koopman operator.
    Section 2.2, Eqs. (30)–(36) assert this correspondence; the paper lists it as an assumption ('Locatable invariant sub-set estimates exist') rather than proving it.
  • ad hoc to paper The observable g(x)=x is sufficient for highly nonlinear dynamics provided regions are partitioned finely enough.
    Section 2.2, assumption 2. The authors' own Lorenz results (Section 3.2.2) show that with g(x)=x the spectra remain high-dimensional, so this assumption is questionable.
  • ad hoc to paper The topological transform ϕ(·) and its inverse preserve the dynamical and spectral content of the local region.
    Section 2.6.3 introduces row-wise interpolation to stretch non-rectangular regions to rectangular form; no proof is given that DMD eigenvalues on the transformed data describe the original evolution.
  • domain assumption Measurement error is IID normal with covariance σ^2 W and can be removed by SVDe whitening.
    Section 2.8 defines the error model; this is a standard but nontrivial assumption for noisy real-world data.
  • standard math Takens' embedding theorem justifies the Hankel matrix construction for the Lorenz system.
    Section 3.2 uses time-delay embeddings to form a 903×700 Hankel matrix; the paper cites Takens via [26, 3].
invented entities (1)
  • Invariant decomposition and its partition categories (I, cS, G, pO)
    purpose: Conceptual framework asserting that global dynamics can be represented as a disjoint union of locally invariant Koopman regions, including approximations of full orbits, continuous spectra, non-straight interfaces, and partial orbits.
    Introduced in Eqs. (30)–(36) and (48) as a guiding principle, but no independent test demonstrates that the learned fuzzy regions are dynamically invariant.

how reviews work

0 comments
Cite this review

Pith. "Pith review of fSRD: Fuzzy Spectral Region Decomposition -- Automated Multi Operator Koopman Representations via an Adaptive Spectral Learning Architecture." pith.science (2026). https://pith.science/paper/2KZTRSI5

@misc{pith2026260717990,
  author       = {Pith},
  title        = {Pith review of: fSRD: Fuzzy Spectral Region Decomposition -- Automated Multi Operator Koopman Representations via an Adaptive Spectral Learning Architecture},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2KZTRSI5}},
  note         = {Machine review of arXiv:2607.17990}
}
read the original abstract

Highly nonlinear chaotic dynamical systems remain difficult to model due to fundamental trade-offs between complexity, expressivity, and data efficiency. Modern machine learning methods achieve strong predictive performance but often rely on a-priori system knowledge or curated data with limited interpretability. Koopman operator theory offers a promising direction via linear representation in an infinite-dimensional observable space. However, many data-driven Koopman methods seek globally valid operators for which useful finite-dimensional spectral embeddings remain difficult to identify under these constraints. To overcome associated limitations, we introduce Fuzzy Spectral Region Decomposition (fSRD), a fully automated learning framework for estimating finite Koopman representation via multiple operators. The proposed method realizes a data-adaptive framework for assembling locally invariant embeddings, termed Invariant Decomposition. fSRD achieves highly accurate linear reconstructions of nonlinear systems while learning finite-dimensional representations of their induced evolution operators, bridging interpretable operator-theoretic models with expressive data-driven sequence learning. These embeddings are adaptively constructed via a global fuzzy tree model, drawing inspiration from fuzzy neural architectures to learn the induced dynamics while prioritizing parsimonious solutions. Empirical results across canonical chaotic systems (e.g., Lorenz and Duffing) and high-dimensional real-world data demonstrate strong predictive accuracy, interpretability, and robust expressivity across data-rich and data-limited regimes, highlighting the method's generality.

Figures

Figures reproduced from arXiv: 2607.17990 by the authors.

Figure 1
Figure 1. Left - Non-straight invariant interface example (black dashed line = interface, [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗
Figure 2
Figure 2. (a) fSRD network architecture. (b) Corresponding forward pass hierarchal tree [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. (a) Example split in top-down premise space. (b) & (c) local [PITH_FULL_IMAGE:figures/full_fig_p021_3.png] view at source ↗
Figures from the paper (27 more)
Figure 4
Figure 4. Figure 4: : SVD alignment issues corrected via rotational transform. Adapted from [16]. [PITH_FULL_IMAGE:figures/full_fig_p022_4.png]
Figure 5
Figure 5. Figure 5: The topological transform ϑ(·) applied to three typical true regions of interest examples given the use of sigmoidal axis oblique splits available. SM-11 provides an example algorithm for ϑ(·) utilizing linear interpolation [65] for speed and simplicity. Further discus…
Figure 6
Figure 6. Figure 6: : Comparison of resultant spectra (e), when applying SVD with the topological [PITH_FULL_IMAGE:figures/full_fig_p025_6.png]
Figure 7
Figure 7. Figure 7: : Full procedure for (a) assigning local co-ordinates and applying a topological [PITH_FULL_IMAGE:figures/full_fig_p026_7.png]
Figure 8
Figure 8. Figure 8: : (a) Duffing attractor trajectories with equally spaced sampling. (b) Duffing [PITH_FULL_IMAGE:figures/full_fig_p031_8.png]
Figure 9
Figure 9. Figure 9: DMD with r = 1e − 4 truncation fitted Duffing equally spaced sampling data • Sampling symmetry, regardless of sampling method, can systematically bias results if the sampling pattern aligns with inherent system structure (e.g., resonance, periodic￾ity, orbital symmetry…
Figure 10
Figure 10. Figure 10: DMD with r = 1e − 4 truncation on Duffing random uniform sampling data (i.e., 69% accuracy). This is despite the corresponding spectra containing vastly more eigenvalues/parameters as in figure-10-(d) (40 → 64). The demonstrated reliance on systems priors’ limits EDMD…
Figure 11
Figure 11. Figure 11: : fSRD on Duffing random uniform sampling data. Global NRMSE = 0.026 [PITH_FULL_IMAGE:figures/full_fig_p034_11.png]
Figure 12
Figure 12. Figure 12: : fSRD model in original state-space. Black lines = fuzzy interpolation. [PITH_FULL_IMAGE:figures/full_fig_p034_12.png]
Figure 13
Figure 13. Figure 13: Bootstrap sampled 2k factorial DOE. (a) spatial design, (b) temporal design Consider the two-factor factorial design of experiments (DOE) [78] presented in figure￾13. The Duffing system is sampled via Monte Carlo with replacement (i.e., bootstrap sampled) 10 times for…
Figure 14
Figure 14. Figure 14: Factorial Experiment results (fSRD) Several trends and comparisons are observed, detailed within SM-27. Summarising, fSRD reliably characterises highly accurate, precise and non-overfitting models of the Duff￾ing system independent of noise, trajectory position, exper…
Figure 15
Figure 15. Figure 15: : (a) Lorenz system bifurcation graph with [PITH_FULL_IMAGE:figures/full_fig_p037_15.png]
Figure 16
Figure 16. Figure 16: DMD (r = 1e−4) on Lorenz hankel matrix. Note Residual = 1 ⇒ Err = 100% example, the much longer signal length here may reduce the ability for a singular set of eigenfunctions to permeate the entire signal on two fronts, both by limited single opera￾tor representation …
Figure 17
Figure 17. Figure 17: : fSRD applied to Lorenz hankel matrix [PITH_FULL_IMAGE:figures/full_fig_p040_17.png]
Figure 18
Figure 18. Figure 18: : fSRD model of hankel Lorenz in state space with per region breakdown [PITH_FULL_IMAGE:figures/full_fig_p040_18.png]
Figure 19
Figure 19. Figure 19: : Region contributions to state space interpolated signal, col’s 375-650 of (55) [PITH_FULL_IMAGE:figures/full_fig_p041_19.png]
Figure 20
Figure 20. Figure 20: : SVD Spectra for each region of hankel Lorenz fSRD model [PITH_FULL_IMAGE:figures/full_fig_p042_20.png]
Figure 21
Figure 21. Figure 21: Bifurcation graph of Lorenz along ρ. (c) corresponds to (58), while (d)-(f) highlight shifts in global behaviour The limitation in local dimensionality shown for systems with pure continuous spectral dynamics can be readily tackled via the use of co-ordinate transform…
Figure 22
Figure 22. Figure 22: DMD (r = 1e − 4) on Lorenz bifurcation region (58) [PITH_FULL_IMAGE:figures/full_fig_p045_22.png]
Figure 23
Figure 23. Figure 23: fSRD (r = 1e − 4) on Lorenz bifurcation region (58) 45 [PITH_FULL_IMAGE:figures/full_fig_p045_23.png]
Figure 24
Figure 24. Figure 24: : fSRD’s Lorenz bifurcation model compared with example state-space attractors [PITH_FULL_IMAGE:figures/full_fig_p046_24.png]
Figure 25
Figure 25. Figure 25: fSRD model regions as ρ increases, highlighting interpretability. Light to darker colour transition corresponds to ρ’s increase in each region Figure-25 demonstrates via visual graphics how these local regions can be used for system interpretation. Each column of the …
Figure 26
Figure 26. Figure 26: : Eigenvalue spectra of fSRD model per region [PITH_FULL_IMAGE:figures/full_fig_p048_26.png]
Figure 27
Figure 27. Figure 27: : Comparison with first two modes of each region (colours darken as [PITH_FULL_IMAGE:figures/full_fig_p048_27.png]
Figure 28
Figure 28. Figure 28: Solar data ”Mother’s Day Storm”, (a) example and (b) snapshot matrix [PITH_FULL_IMAGE:figures/full_fig_p050_28.png]
Figure 29
Figure 29. Figure 29: fSRD on Solar data, (a) convergance and (b) pruning results [PITH_FULL_IMAGE:figures/full_fig_p051_29.png]
Figure 30
Figure 30. Figure 30: (a) DMD and (b) fSRD residuals on Solar data [PITH_FULL_IMAGE:figures/full_fig_p052_30.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

125 extracted references · 19 canonical work pages

  1. [15]

    Bokor.fSRD: Fuzzy Spectral Region Decomposition – Automated Multi Operator Koopman Representations via an Adaptive Spectral Learning Architecture

    C. Bokor.fSRD: Fuzzy Spectral Region Decomposition – Automated Multi Operator Koopman Representations via an Adaptive Spectral Learning Architecture. PhD thesis, Oxford Brookes University, 2026. submitted, availability pending

    2026

  2. [1]

    A. A. Adeniran and S. E. Ferik. Modeling and identification of nonlinear systems: A review of the multimodel approach—part 1.IEEE Transactions on Systems, Man, and Cybernetics: Systems, 47:1149–1159, 7 2017. ISSN 2168-2216. doi: 10.1109/ TSMC.2016.2560147. URLhttp://ieeexplore.ieee.org/document/7469784/

    arXiv 2017

  3. [2]

    Alzubaidi, J

    L. Alzubaidi, J. Zhang, A. J. Humaidi, A. Al-Dujaili, Y. Duan, O. Al-Shamma, J. San- tamar ´ ıa, M. A. Fadhel, M. Al-Amidie, and L. Farhan. Review of deep learning: concepts, cnn architectures, challenges, applications, future directions.Journal of Big Data, 8:53, 3 2021. ISSN 2196-1115. doi: 10.1186/s40537-021-00444-8. URL https://doi.org/10.1186/s40537-...

    doi:10.1186/s40537-021-00444-8 2021

  4. [3]

    Arbabi and I

    H. Arbabi and I. Mezi´ c. Ergodic theory, dynamic mode decomposition, and compu- tation of spectral properties of the koopman operator.SIAM Journal on Applied Dy- namical Systems, 16:2096–2126, 1 2017. ISSN 1536-0040. doi: 10.1137/17M1125236

    doi:10.1137/17m1125236

  5. [4]

    Babuˇ ska and H

    R. Babuˇ ska and H. Verbruggen. Neuro-fuzzy methods for nonlinear system identifi- cation.Annual Reviews in Control, 27:73–85, 1 2003. ISSN 13675788. doi: 10.1016/ S1367-5788(03)00009-9. URLhttps://linkinghub.elsevier.com/retrieve/pii/ S1367578803000099

    2003

  6. [5]

    S. Bagheri. Effects of weak noise on oscillating flows: Linking quality factor, floquet modes, and koopman spectrum.Physics of Fluids, 26, 9 2014. ISSN 1070-6631. doi: 10.1063/1.4895898

    doi:10.1063/1.4895898 2014

  7. [6]

    B¨ anfer and O

    O. B¨ anfer and O. Nelles. Polynomial model tree (polymot) — a new training algorithm for local model networks with higher degree polynomials. In2009 IEEE International Conference on Control and Automation, pages 1571–1576. IEEE, 12 2009. ISBN 978- 1-4244-4706-0. doi: 10.1109/ICCA.2009.5410547. URLhttp://ieeexplore.ieee. org/document/5410547/

    arXiv 2009

  8. [7]

    B¨ anfer, M

    O. B¨ anfer, M. Franke, and O. Nelles. Adaptive local model networks with higher degree polynomials. InICCAS 2010, pages 168–171. IEEE, 10 2010. ISBN 978-1- 4244-7453-0. doi: 10.1109/ICCAS.2010.5669893

    arXiv 2010

Show all 125 references
  1. [8]

    B¨ anfer, B

    O. B¨ anfer, B. Hartmann, and O. Nelles. Comparison of different subset selection algorithms for learning local model networks with higher degree polynomials. In2010 11th International Conference on Control Automation Robotics & Vision, pages 30–

  2. [9]

    B¨ anfer, B

    O. B¨ anfer, B. Hartmann, and O. Nelles. Polymot versus hilomot - a comparison of two different training algorithms for local model networks.IF AC Proceedings Volumes, 45: 1569–1574, 7 2012. ISSN 14746670. doi: 10.3182/20120711-3-BE-2027.00354

    2012

  3. [10]

    W. T. Barnes, M. G. Bobra, S. D. Christe, N. Freij, L. A. Hayes, J. Ireland, S. Mum- ford, D. Perez-Suarez, D. F. Ryan, A. Y. Shih, P. Chanda, K. Glogowski, R. Hewett, V. K. Hughitt, A. Hill, K. Hiware, A. Inglis, M. S. F. Kirk, S. Konge, J. P. Mason, S. A. Maloney, S. A. Murr...

    2020

  4. [11]

    N. S. Bedrossian.Nonlinear control using linearizing transformations. PhD thesis, Massachusetts Institute of Technology, 1991

    1991

  5. [12]

    A. O. Benz. Flare observations.Living Reviews in Solar Physics, 14:2, 12 2017. ISSN 2367-3648. doi: 10.1007/s41116-016-0004-3

    2017 doi

  6. [13]

    C. M. Bishop.Pattern Recognition and Machine Learning. Springer, 2006. ISBN 978-0387310732

    2006

  7. [14]

    C. Bokor. Solar anomaly detector, 2026. URLhttps://github.com/cbokor/ solar-anomaly-detector. GitHub repository

    2026

  8. [16]

    S. L. Brunton and J. N. Kutz.Data-Driven Science and Engineering: Machine Learn- ing, Dynamical Systems, and Control. Cambridge University Press, 1 2019. ISBN 9781108380690. doi: 10.1017/9781108380690

    2019 doi

  9. [17]

    S. L. Brunton, B. W. Brunton, J. L. Proctor, and J. N. Kutz. Koopman invariant subspaces and finite linear representations of nonlinear dynamical systems for control. PLOS ONE, 10 2016. doi: 10.1371/journal.pone.0150171. URLhttp://dx.doi.org/ 10.1371/journal.pone.0150171

    2016 doi

  10. [18]

    S. L. Brunton, B. W. Brunton, J. L. Proctor, E. Kaiser, and J. N. Kutz. Chaos as an intermittently forced linear system.Nature Communications, 19, 2017. doi: 10.1038/ s41467-017-00030-8. URLhttp://dx.doi.org/10.1038/s41467-017-00030-8

    2017 doi

  11. [19]

    S. L. Brunton, M. Budiˇ si´ c, E. Kaiser, and J. N. Kutz. Modern koopman theory for dynamical systems, 2021. URLhttp://arxiv.org/abs/2102.12086

    2021 arXiv

  12. [20]

    Budiˇ si´ c and I

    M. Budiˇ si´ c and I. Mezi´ c. Geometry of the ergodic quotient reveals coherent struc- tures in flows.Physica D: Nonlinear Phenomena, 241(15):1255–1269, 2012. ISSN 0167-2789. doi: https://doi.org/10.1016/j.physd.2012.04.006. URLhttps://www. sciencedirect.com/science/article/p...

    2012 doi

  13. [21]

    Budiˇ si´ c, R

    M. Budiˇ si´ c, R. Mohr, and I. Mezi´ c. Applied koopmanism.Chaos: An Interdisciplinary Journal of Nonlinear Science, 22, 12 2012. ISSN 1054-1500. doi: 10.1063/1.4772195

    2012 doi

  14. [22]

    Cary and C

    M. Cary and C. Bokor. Regularised iterative generalised least squares with optimal selection of the hyper-parameter for identifying nonlinear phenomenological models. arXiv preprint, 2026. 58 fSRD: Fuzzy Spectral Region Decomposition

    2026

  15. [23]

    Cary and C

    M. Cary and C. Bokor. A repeated measurements approach to soh battery modelling of cyclic aged data in a laboratory environment.arXiv preprint, 2026

    2026

  16. [24]

    Chai and R

    J. Chai and R. G. Sanfelice. On notions and sufficient conditions for forward invariance of sets for hybrid dynamical systems. In2015 54th IEEE Conference on Decision and Control (CDC), pages 2869–2874. IEEE, 12 2015. doi: 10.1109/CDC.2015.7402652

    2015

  17. [25]

    Champion, B

    K. Champion, B. Lusch, J. N. Kutz, and S. L. Brunton. Data-driven discovery of coordinates and governing equations.Proceedings of the National Academy of Sciences of the United States of America, 116:22445–22451, 2019. ISSN 10916490. doi: 10.1073/ pnas.1906995116

    2019

  18. [26]

    M. J. Colbrook. The multiverse of dynamic mode decomposition algorithms, 11 2023. URLhttp://arxiv.org/abs/2312.00137

    2023 arXiv

  19. [27]

    M. J. Colbrook. The mpedmd algorithm for data-driven computations of measure- preserving dynamical systems.SIAM Journal on Numerical Analysis, 61:1585–1608, 6 2023. ISSN 0036-1429. doi: 10.1137/22M1521407

    2023 doi

  20. [28]

    Colonius and W

    F. Colonius and W. Kliemann.The Dynamics of Control. Birkh¨ auser Boston,

  21. [29]

    G. Cybenko. Approximation by superpositions of a sigmoidal function.Mathematics of Control, Signals, and Systems, 2:303–314, 12 1989. ISSN 0932-4194. doi: 10.1007/ BF02551274. URLhttp://link.springer.com/10.1007/BF02551274

    1989 doi

  22. [30]

    A. C. Davison and D. V. Hinkley.Bootstrap Methods and their Application. Cambridge University Press, 10 1997. ISBN 9780521573917. doi: 10.1017/CBO9780511802843

    1997 doi

  23. [31]

    S. T. M. Dawson, M. S. Hemati, M. O. Williams, and C. W. Rowley. Characterizing and correcting for the effect of sensor noise in the dynamic mode decomposition.Ex- periments in Fluids, 57:42, 3 2016. ISSN 0723-4864. doi: 10.1007/s00348-016-2127-7

    2016 doi

  24. [32]

    de la Madrid

    R. de la Madrid. The role of the rigged hilbert space in quantum mechanics.European Journal of Physics, 26:287–312, 4 2005. ISSN 0143-0807. doi: 10.1088/0143-0807/26/ 2/008

    2005 doi

  25. [33]

    Duffing.Erzwungene Schwingungen bei ver¨ anderlicher Eigenfrequenz und ihre tech- nische Bedeutung

    G. Duffing.Erzwungene Schwingungen bei ver¨ anderlicher Eigenfrequenz und ihre tech- nische Bedeutung. Vieweg & Sohn, 1918

    1918

  26. [34]

    B. P. Epps and E. M. Krivitzky. Singular value decomposition of noisy data: mode corruption.Experiments in Fluids, 60:121, 8 2019. ISSN 0723-4864. doi: 10.1007/ s00348-019-2761-y

    2019

  27. [35]

    ISBN 978-1-4244-7814-9

    IEEE, 12 2010. ISBN 978-1-4244-7814-9. doi: 10.1109/ICARCV.2010.5707393

    2010

  28. [36]

    Eykhoff and K

    P. Eykhoff and K. Astrom. System identification - a survey.Automatica, 7:123–162, 1971

    1971

  29. [37]

    B. P. Epps and E. M. Krivitzky. Singular value decomposition of noisy data: noise filtering.Experiments in Fluids, 60:126, 8 2019. ISSN 0723-4864. doi: 10.1007/ s00348-019-2768-4. 59 Bokor et al

    2019

  30. [38]

    Ghosh and J

    R. Ghosh and J. Marecek. Iterated function systems: A comprehensive survey, 11

  31. [39]

    Folkestad, D

    C. Folkestad, D. Pastor, I. Mezi´ c, R. Mohr, M. Fonoberova, and J. Burdick. Extended dynamic mode decomposition with learned koopman eigenfunctions for prediction and control. In2020 American Control Conference (ACC), pages 3906–3913. IEEE, 7 2020. ISBN 978-1-5386-8266-1. doi...

    2020

  32. [40]

    G. H. Golub and C. F. V. Loan.Matrix Computations. Johns Hopkins University Press, 4th edition, 2013

    2013

  33. [41]

    Goodfellow, Y

    I. Goodfellow, Y. Bengio, and A. Courville.Deep Learning. MIT Press,

  34. [42]

    Gohel, P

    P. Gohel, P. Singh, and M. Mohanty. Explainable ai: current status and future directions, 7 2021. URLhttp://arxiv.org/abs/2107.07045

    2021 arXiv

  35. [43]

    D. Grobman. Homeomorphisms of systems of differential equations.Doklady Akademii Nauk SSSR, 128:880–881, 1959

    1959

  36. [44]

    C. Gros. Pushing the complexity barrier: diminishing returns in the sciences, 9 2012. URLhttp://arxiv.org/abs/1209.2725

    2012 arXiv

  37. [45]

    Guckenheimer and P

    J. Guckenheimer and P. Holmes.Nonlinear Oscillations, Dynamical Systems, and Bifurcations of Vector Fields, volume 42. Springer New York, 1983. ISBN 978-1- 4612-7020-1. doi: 10.1007/978-1-4612-1140-2. URLhttp://link.springer.com/ 10.1007/978-1-4612-1140-2

    1983 doi

  38. [46]

    Goutte.Statistical learning and regularisation for regression

    C. Goutte.Statistical learning and regularisation for regression. PhD thesis, Munich, 1997

    1997

  39. [47]

    P. Hartman. A lemma in the theory of structural stability of differential equations.Proceedings of the American Mathematical Society, 11:610, 8 1960. ISSN 00029939. doi: 10.2307/2034720. URLhttps://doi.org/10.1090/ S0002-9939-1960-0121542-7

    1960 doi

  40. [48]

    Hartman.Ordinary Differential Equations

    P. Hartman.Ordinary Differential Equations. Society for Industrial and Applied Mathematics, 1 2002. ISBN 978-0-89871-510-1. doi: 10.1137/1.9780898719222. URL http://epubs.siam.org/doi/book/10.1137/1.9780898719222

    2002 doi

  41. [49]

    Hartmann and O

    B. Hartmann and O. Nelles. Structure trade-off strategy for local model networks. In 2012 IEEE International Conference on Control Applications, pages 451–456. IEEE, 10 2012. ISBN 978-1-4673-4505-7. doi: 10.1109/CCA.2012.6402404. URLhttp: //ieeexplore.ieee.org/document/6402404...

    2012

  42. [50]

    Hall.Quantum Theory for Mathematicians

    B. Hall.Quantum Theory for Mathematicians. Springer, 2013. ISBN 9781461471158

    2013

  43. [51]

    Hoffman and R

    K. Hoffman and R. Kunze.Linear Algebra. Pearson, 2nd edition, 1971. ISBN 978- 0135367971

    1971

  44. [52]

    Horn and C

    R. Horn and C. Johnson.Matrix Analysis. Cambridge University Press, 2nd edition,

  45. [53]

    K. Hornik. Approximation capabilities of multilayer feedforward networks.Neural Networks, 4:251–257, 1991. ISSN 08936080. doi: 10.1016/0893-6080(91)90009-T

    1991 doi

  46. [54]

    Hartmann, O

    B. Hartmann, O. Nelles, I. ˇSkrjanc, and A. Sodja. Supervised hierarchical clustering (suhiclust) for nonlinear system identification. In2009 IEEE Symposium on Com- putational Intelligence in Control and Automation, CICA 2009 - Proceedings, pages 41–48, 2009. ISBN 978142442752...

    2009

  47. [55]

    Johnson-Groh

    M. Johnson-Groh. How nasa tracked the most intense solar storm in decades, 5 2024. URLhttps://science.nasa.gov/science-research/heliophysics/ how-nasa-tracked-the-most-intense-solar-storm-in-decades/

    2024

  48. [56]

    M. R. Jovanovi´ c, P. J. Schmid, and J. W. Nichols. Sparsity-promoting dynamic mode decomposition.Physics of Fluids, 26, 2 2014. ISSN 1070-6631. doi: 10.1063/1.4863670

    2014 doi

  49. [57]

    M. Kamb, E. Kaiser, S. L. Brunton, and J. N. Kutz. Time-delay observables for koopman: Theory and applications.SIAM Journal on Applied Dynamical Systems, 19:886–917, 2020. ISSN 15360040. doi: 10.1137/18M1216572

    2020 doi

  50. [58]

    Y. Jin, L. Hou, and S. Zhong. Extended dynamic mode decomposition with invertible dictionary learning.Neural Networks, 173:106177, 5 2024. ISSN 08936080. doi: 10.1016/j.neunet.2024.106177

    2024

  51. [59]

    B. O. Koopman. Hamiltonian systems and transformation in hilbert space.Pro- ceedings of the National Academy of Sciences, 17:315–318, 5 1931. ISSN 0027-8424. doi: 10.1073/pnas.17.5.315. URLhttps://pnas.org/doi/full/10.1073/pnas.17. 5.315

    1931 doi

  52. [60]

    Korda and I

    M. Korda and I. Mezi´ c. Optimal construction of koopman eigenfunctions for prediction and control, 10 2018. URLhttp://arxiv.org/abs/1810.08733

    2018 arXiv

  53. [61]

    Korda and I

    M. Korda and I. Mezi´ c. On convergence of extended dynamic mode decomposition to the koopman operator.Journal of Nonlinear Science, 28:687–710, 4 2018. ISSN 0938-8974. doi: 10.1007/s00332-017-9423-0

    2018 doi

  54. [62]

    S. Klus, P. Koltai, and C. Sch¨ utte. On the numerical approximation of the perron- frobenius and koopman operator.Journal of Computational Dynamics, 12 2015. doi: 10.3934/jcd.2016003. URLhttp://dx.doi.org/10.3934/jcd.2016003

    2015 doi

  55. [63]

    Lang.Linear Algebra

    S. Lang.Linear Algebra. Springer New York, 1987. ISBN 978-1-4419-3081-

    1987

  56. [64]

    J. R. Lemen, A. M. Title, D. J. Akin, P. F. Boerner, C. Chou, J. F. Drake, D. W. Duncan, C. G. Edwards, F. M. Friedlaender, G. F. Heyman, N. E. Hurlburt, N. L. Katz, G. D. Kushner, M. Levay, R. W. Lindgren, D. P. Mathur, E. L. McFeaters, S. Mitchell, R. A. Rehse, C. J. Schrijv...

  57. [65]

    Lepot, J.-B

    M. Lepot, J.-B. Aubin, and F. Clemens. Interpolation in time series: An introductive overview of existing methods, their performance criteria and uncertainty assessment. Water, 9:796, 10 2017. ISSN 2073-4441. doi: 10.3390/w9100796

    2017 doi

  58. [66]

    Lan and I

    Y. Lan and I. Mezi´ c. Linearization in the large of nonlinear systems and koopman operator spectrum.Physica D: Nonlinear Phenomena, 242:42–53, 1 2013. ISSN 01672789. doi: 10.1016/j.physd.2012.08.017. 61 Bokor et al

    2013 doi

  59. [67]

    L. Liberti. Introduction to global optimization. 2006. URLhttps://api. semanticscholar.org/CorpusID:14115596

    2006

  60. [68]

    URLhttp://link.springer.com/10.1007/ 978-1-4757-1949-9

    doi: 10.1007/978-1-4757-1949-9. URLhttp://link.springer.com/10.1007/ 978-1-4757-1949-9

    1949 doi

  61. [69]

    E. N. Lorenz. Deterministic nonperiodic flow.Journal of the Atmospheric Sciences, 20:130–141, 1963. doi: https://doi.org/10.1175/1520-0469(1963)020%3C0130:DNF% 3E2.0.CO;2. URLhttps://doi.org/10.1175/1520-0469(1963)020%3C0130:DNF% 3E2.0.CO;2

    1963 doi

  62. [70]

    Lusch, J

    B. Lusch, J. N. Kutz, and S. L. Brunton. Deep learning for universal linear embeddings of nonlinear dynamics.Nature Communications, 9, 12 2018. ISSN 20411723. doi: 10.1038/s41467-018-07210-0

    2018 doi

  63. [71]

    Mahfuz, M

    A. Mahfuz, M. A. Mannan, and S. M. Muyeen. Enhancement of the hilomot algorithm with modified em and modified pso algorithms for nonlinear systems identification. Electronics, 11:729, 2 2022. ISSN 2079-9292. doi: 10.3390/electronics11050729

    2022 doi

  64. [72]

    X. Li, H. Xiong, X. Li, X. Wu, X. Zhang, J. Liu, J. Bian, and D. Dou. Interpretable deep learning: Interpretation, interpretability, trustworthiness, and beyond, 3 2021. URLhttp://arxiv.org/abs/2103.10689

    2021 arXiv

  65. [73]

    I. Mezi´ c. Spectral properties of dynamical systems, model reduction and de- compositions.Nonlinear Dynamics, 41:309–325, 8 2005. ISSN 0924090X. doi: 10.1007/s11071-005-2824-x. 62 fSRD: Fuzzy Spectral Region Decomposition

    2005 doi

  66. [74]

    Ljung.System Identification: Theory for the User

    L. Ljung.System Identification: Theory for the User. Pearson Education Limited, 2nd edition, 1999. ISBN 978-0136566953

    1999

  67. [75]

    I. Mezi´ c. Spectrum of the koopman operator, spectral expansions in functional spaces, and state space geometry, 2 2017. URLhttp://arxiv.org/abs/1702.07597

    2017 arXiv

  68. [76]

    Mezi´ c and A

    I. Mezi´ c and A. Banaszuk. Comparison of systems with complex behavior.Physica D: Nonlinear Phenomena, 197:101–133, 10 2004. ISSN 01672789. doi: 10.1016/j.physd. 2004.06.015

    2004 doi

  69. [77]

    Mezi´ c and S

    I. Mezi´ c and S. Wiggins. A method for visualization of invariant sets of dynami- cal systems based on the ergodic partition.Chaos: An Interdisciplinary Journal of Nonlinear Science, 9(1):213–218, 3 1999. doi: 10.1063/1.166399

    1999 doi

  70. [78]

    J. D. Meiss.Differential Dynamical Systems. Society for Industrial and Applied Mathematics, 1 2007. ISBN 978-0-89871-635-1. doi: 10.1137/1.9780898718232. URL http://epubs.siam.org/doi/book/10.1137/1.9780898718232

    2007 doi

  71. [79]

    D. C. Montgomery, E. A. Peck, and G. G. Vining.Introduction to Linear Regression Analysis. Wiley, 5th edition, 2012

    2012

  72. [80]

    I. Mezi´ c. Analysis of fluid flows via spectral properties of the koopman opera- tor.Annual Review of Fluid Mechanics, 45:357–378, 1 2013. ISSN 0066-4189. doi: 10.1146/annurev-fluid-011212-140652. URLhttps://www.annualreviews.org/ doi/10.1146/annurev-fluid-011212-140652

    2013 doi

  73. [81]

    S. P. Nandanoori, S. Sinha, and E. Yeung. Data-driven operator theoretic methods for phase space learning and analysis, 6 2021. URLhttp://arxiv.org/abs/2106.15678

    2021 arXiv

  74. [82]

    O. Nelles. Local basis function networks for identification of a turbocharger. In UKACC International Conference on Control. Control ’96, volume 1996, pages 7–

    1996

  75. [83]

    Nelles.Nonlinear System Identification

    O. Nelles.Nonlinear System Identification. Springer Berlin Heidelberg, 2001. ISBN 978-3-642-08674-8. doi: 10.1007/978-3-662-04323-3. URLhttp://link.springer. com/10.1007/978-3-662-04323-3

    2001 doi

  76. [84]

    D. C. Montgomery.Design and Analysis of Experiments. Wiley, 2019. ISBN 978-1-119-49244-3. URLhttps://www.wiley.com/en-us/Design+and+Analysis+ of+Experiments%2C+10th+Edition-p-9781119492443

    2019

  77. [85]

    Niederreiter.Random Number Generation and Quasi-Monte Carlo Methods

    H. Niederreiter.Random Number Generation and Quasi-Monte Carlo Methods. So- ciety for Industrial and Applied Mathematics, 1 1992. ISBN 978-0-89871-295-7. doi: 10.1137/1.9781611970081

    1992 doi

  78. [86]

    S. P. Nandanoori, S. Sinha, and E. Yeung. Data-driven operator theoretic methods for global phase space learning, 10 2019. URLhttp://arxiv.org/abs/1910.03011

    2019 arXiv

  79. [87]

    Ott.Chaos in Dynamical Systems

    E. Ott.Chaos in Dynamical Systems. Cambridge University Press, 8 2002. ISBN 9780521811965. doi: 10.1017/CBO9780511803260

    2002 doi

  80. [88]

    E. Ott, C. Grebogi, and J. A. Yorke. Controlling chaos.Physical Review Letters, 64: 1196–1199, 3 1990. ISSN 0031-9007. doi: 10.1103/PhysRevLett.64.1196

    1990 doi

  81. [89]

    ISBN 0 85296 666 0

    IEE, 1996. ISBN 0 85296 666 0. doi: 10.1049/cp:19960518. URLhttps:// ieeexplore.ieee.org/document/651344

    1996 doi

  82. [90]

    Queff´ elec.Substitution Dynamical Systems - Spectral Analysis, volume 1294

    M. Queff´ elec.Substitution Dynamical Systems - Spectral Analysis, volume 1294. Springer Berlin Heidelberg, 2010. ISBN 978-3-642-11211-9. doi: 10.1007/ 978-3-642-11212-6

    2010

  83. [91]

    O. Nelles. Axes-oblique partitioning strategies for local model networks. InIEEE International Symposium on Intelligent Control, pages 2378–2383. IEEE, 10 2006. ISBN 0-7803-9798-3. doi: 10.1109/ISIC.2006.285656. URLhttps://ieeexplore. ieee.org/document/4064981

    2006

  84. [92]

    C. W. Rowley, I. Mezi´ c, S. Bagheri, P. Schlatter, and D. S. Henningson. Spectral analysis of nonlinear flows.Journal of Fluid Mechanics, 641:115–127, 2009. ISSN 14697645. doi: 10.1017/S0022112009992059

    2009 doi

  85. [93]

    A. V. Oppenheim, R. W. Schafer, and J. R. Buck.Discrete-Time Signal Processing. Pearson, 3rd edition, 1998. 63 Bokor et al

    1998

  86. [94]

    E. N. Sanchez and A. Y. Alanis. Adaptive approximation based control: Unifying neural, fuzzy and traditional adaptive approximation approaches (farrell, j.a. and polycarpou, m.m. [book review].IEEE Transactions on Neural Networks, 19:731– 732, 4 2008. ISSN 1045-9227. doi: 10.1...

    2008

  87. [95]

    I. H. Sarker. Machine learning: Algorithms, real-world applications and re- search directions.SN Computer Science, 2:160, 5 2021. ISSN 2662-995X. doi: 10.1007/s42979-021-00592-x. URLhttps://link.springer.com/10.1007/ s42979-021-00592-x

    2021 doi

  88. [96]

    J. L. Proctor, S. L. Brunton, and J. N. Kutz. Dynamic mode decomposition with con- trol.SIAM Journal on Applied Dynamical Systems, 15:142–161, 2016. ISSN 15360040. doi: 10.1137/15M1013857

    2016 doi

  89. [97]

    Schm¨ olter and J

    E. Schm¨ olter and J. Berdermann. The impact of the 2024 mother’s day storm on aircraft surveillance across europe.Space Weather, 23, 12 2025. ISSN 1542-7390. doi: 10.1029/2025SW004718

    2024 doi

  90. [98]

    L. M. Rios and N. V. Sahinidis. Derivative-free optimization: a review of algorithms and comparison of software implementations.Journal of Global Optimization, 56: 1247–1293, 7 2013. ISSN 0925-5001. doi: 10.1007/s10898-012-9951-y

    2013 doi

  91. [99]

    Searle and A

    S. Searle and A. Khuri.Matrix Algebra Useful for Statistics. Wiley, 2nd edition, 2017. ISBN 978-1-118-93514-9. 64 fSRD: Fuzzy Spectral Region Decomposition

    2017

  92. [100]

    Salova, J

    A. Salova, J. Emenheiser, A. Rupe, J. P. Crutchfield, and R. M. D’Souza. Koopman operator and its approximations for systems with symmetries.Chaos, 29, 9 2019. ISSN 10541500. doi: 10.1063/1.5099091

    2019 doi

  93. [101]

    Sinha, S

    S. Sinha, S. P. Nandanoori, and E. Yeung. Koopman operator methods for global phase space exploration of equivariant dynamical systems. InIF AC-PapersOnLine, volume 53, pages 1150–1155. Elsevier B.V., 2020. doi: 10.1016/j.ifacol.2020.12.1322

    2020 doi

  94. [102]

    Sparrow.The Lorenz Equations: Bifurcations, Chaos, and Strange Attractors, volume 41

    C. Sparrow.The Lorenz Equations: Bifurcations, Chaos, and Strange Attractors, volume 41. Springer New York, 1982. ISBN 978-0-387-90775-8. doi: 10.1007/ 978-1-4612-5767-7

    1982

  95. [103]

    P. J. SCHMID. Dynamic mode decomposition of numerical and experimental data.Journal of Fluid Mechanics, 656:5–28, 8 2010. ISSN 0022-1120. doi: 10.1017/S0022112010001217. URLhttps://doi.org/10.1017/S0022112010001217

    2010 doi

  96. [104]

    Staffans.Well-Posed Linear Systems

    O. Staffans.Well-Posed Linear Systems. Cambridge University Press, 2 2005. ISBN 9780521825849. doi: 10.1017/CBO9780511543197

    2005 doi

  97. [105]

    G. Schwarz. Estimating the dimension of a model.The Annals of Statistics, 6, 3 1978. ISSN 0090-5364. doi: 10.1214/aos/1176344136

    1978

  98. [106]

    Takagi and M

    T. Takagi and M. Sugeno. Fuzzy identification of systems and its applications to modeling and control.IEEE Transactions on Systems, Man, and Cybernetics, SMC- 15:116–132, 1 1985. ISSN 0018-9472. doi: 10.1109/TSMC.1985.6313399

    1985

  99. [107]

    Shahzad.Pruned hierarchical local model networks for nonlinear system identifi- cation

    A. Shahzad.Pruned hierarchical local model networks for nonlinear system identifi- cation. PhD thesis, Loughborough University, 2020

    2020

  100. [108]

    A. N. Tikhonov and V. Y. Arsenin.Solution of ill-posed problems. D.C.: John Wiley & Sons, 1977

    1977

  101. [109]

    J. H. Tu, C. W. Rowley, D. M. Luchtenburg, S. L. Brunton, and J. N. Kutz. On dynamic mode decomposition: Theory and applications.Journal of Computational Dynamics, 1:391–421, 2014. ISSN 21582505. doi: 10.3934/jcd.2014.1.391. URL https://www.aimsciences.org/article/doi/10.3934/...

    2014 doi

  102. [110]

    J. C. Sprott.Chaos and Time-Series Analysis. Oxford University PressOxford, 1

  103. [111]

    Valentin, A

    A. Valentin, A. Luo, and X. Fu.Nonlinear Dynamics and Complexity, vol- ume 8. Springer International Publishing, 2014. ISBN 978-3-319-02352-6. doi: 10.1007/978-3-319-02353-3. URLhttps://link.springer.com/10.1007/ 978-3-319-02353-3

    2014 doi

  104. [112]

    B. Verd, A. Crombach, and J. Jaeger. Classification of transient behaviours in a time- dependent toggle switch model.BMC Systems Biology, 8, 4 2014. ISSN 17520509. doi: 10.1186/1752-0509-8-43. 65 Bokor et al

    2014 doi

  105. [113]

    N. Sugiura. Further analysis of the data by akaike’s information criterion and the finite corrections.Communications in Statistics - Theory and Methods, 7:13–26, 1

  106. [114]

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

  107. [116]

    A. K. Tangirala.Principles of System Identification. CRC Press, 1st edition,

  108. [120]

    E. Turco. Tools for the numerical solution of inverse problems in structural mechanics: review and research perspectives.European Journal of Environmental and Civil En- gineering, 21:509–554, 5 2017. ISSN 1964-8189. doi: 10.1080/19648189.2015.1134673

    2017

  109. [123]

    M. O. Williams, I. G. Kevrekidis, and C. W. Rowley. A data–driven approximation of the koopman operator: Extending dynamic mode decomposition.Journal of Nonlin- ear Science, 25:1307–1346, 12 2015. ISSN 14321467. doi: 10.1007/s00332-015-9258-5

    2015 doi

  110. [125]

    doi: 10.3934/jcd.2015005

    ISSN 2158-2505. doi: 10.3934/jcd.2015005. 66

    doi

  111. [1978]

    doi: 10.1080/03610927808827599

    ISSN 0361-0926. doi: 10.1080/03610927808827599

    doi

  112. [2000]

    doi: 10.1007/978-1-4612-1350-5

    ISBN 978-1-4612-7107-9. doi: 10.1007/978-1-4612-1350-5. URLhttp: //link.springer.com/10.1007/978-1-4612-1350-5

    doi

  113. [2003]

    doi: 10.1093/oso/9780198508397.001.0001

    ISBN 9780198508397. doi: 10.1093/oso/9780198508397.001.0001

  114. [2012]

    doi: 10.1007/s11207-011-9776-8

    ISSN 0038-0938. doi: 10.1007/s11207-011-9776-8

    doi

  115. [2015]

    doi: 10.1201/9781315222509

    ISBN 9781315222509. doi: 10.1201/9781315222509. URLhttps://www. taylorfrancis.com/books/9781439896020

  116. [2016]

    URLhttps://mitpress.mit.edu/9780262035613/ deep-learning/

    ISBN 9780262035613. URLhttps://mitpress.mit.edu/9780262035613/ deep-learning/

  117. [2022]

    URLhttp://arxiv.org/abs/2211.14661

    arXiv

Pith tools

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