Spatio-Temporal Prediction via Operator-Valued RKHS and Koopman Approximation

arxiv: 2507.12474 · v2 · submitted 2025-07-03 · 🧮 math.GM

Spatio-Temporal Prediction via Operator-Valued RKHS and Koopman Approximation

Mahishanka Withanachchi This is my paper

Pith reviewed 2026-05-19 07:00 UTC · model grok-4.3

classification 🧮 math.GM

keywords operator-valued RKHSKoopman operatorsspatio-temporal predictiondynamical systemsrepresenter theoremsSobolev regularitykernel approximationsspectral convergence

0 comments p. Extension

The pith

Integrating Sobolev regularity with Koopman theory into operator-valued RKHS produces representer theorems and spectral convergence for dynamical system learning.

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

The paper builds a framework for spatio-temporal prediction of time-varying vector fields by combining operator-valued reproducing kernel Hilbert spaces with Koopman operator approximations. It proves new representer theorems for time-aligned interpolation along with Sobolev-based approximation bounds and guarantees on spectral convergence. These foundations support data-driven reduced-order modeling and forecasting in nonlinear dynamical systems. A sympathetic reader would see value in the shift from empirical kernel methods toward rigorously bounded predictions that respect the underlying smoothness of the vector fields.

Core claim

By integrating Sobolev regularity with Koopman operator theory, the authors establish representer theorems, approximation rates, and spectral convergence results for kernel-based learning of dynamical systems using operator-valued reproducing kernel Hilbert spaces for spatio-temporal prediction of time-varying vector fields.

What carries the argument

Operator-valued reproducing kernel Hilbert spaces (OV RKHS) paired with Koopman operator approximations, which together enable time-aligned interpolation and spectral analysis of smooth vector fields.

If this is right

Representer theorems for time-aligned OV RKHS interpolation reduce the prediction task to finite-dimensional linear algebra.
Sobolev approximation bounds quantify how prediction error decreases with increasing smoothness of the vector fields.
Kernel Koopman operator approximations deliver reduced-order models whose spectra converge to the true system spectra.
Spectral convergence guarantees ensure reliable long-term forecasting for complex nonlinear dynamical systems.

Where Pith is reading between the lines

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

The same regularity-based bounds could be tested on hybrid models that combine the kernel framework with neural network approximations of the vector fields.
If the convergence rates hold, they would directly inform sample-complexity requirements for learning tasks in fluid dynamics or climate data.
Extensions to non-Euclidean domains might follow by replacing the underlying Sobolev space with manifold-valued versions while preserving the Koopman spectral structure.

Load-bearing premise

The dynamical systems and time-varying vector fields possess sufficient Sobolev regularity to support the OV RKHS interpolation and kernel Koopman approximations.

What would settle it

A concrete dynamical system whose vector fields lack the assumed Sobolev regularity, for which the derived approximation rates or spectral convergence rates fail to hold in explicit numerical tests.

Figures

Figures reproduced from arXiv: 2507.12474 by Mahishanka Withanachchi.

**Figure 1.** Figure 1: Left: Log-log plot of L 2 error vs. number of data points N. Right: Predicted vs. true vector field. where † denotes the Moore-Penrose pseudoinverse. We compute its leading eigenvalues and eigenfunctions. Evaluation. We analyze: • Spectral convergence: |λ N k − λk| → 0 as N → ∞. • Operator norm decay: ∥KN − K∥op → 0 [PITH_FULL_IMAGE:figures/full_fig_p019_1.png] view at source ↗

**Figure 2.** Figure 2: Left: Convergence of Koopman eigenvalues. Right: Visualized Koopman modes (real and [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗

**Figure 3.** Figure 3: Left: Forecasting error decay as r increases. Right: Koopman spectrum decay showing fast convergence. 6 Related Work The theory of Reproducing Kernel Hilbert Spaces (RKHS) plays a central role in nonparametric statistical learning, with seminal contributions in kernel ridge regression, support vector machines, and Gaussian processes. The classical framework, however, primarily concerns scalar-valued functi… view at source ↗

read the original abstract

We develop a comprehensive framework for spatio-temporal prediction of time-varying vector fields using operator-valued reproducing kernel Hilbert spaces (OV RKHS). By integrating Sobolev regularity with Koopman operator theory, we establish representer theorems, approximation rates, and spectral convergence results for kernel-based learning of dynamical systems. Our theoretical contributions include new representer theorems for time-aligned OV RKHS interpolation, Sobolev approximation bounds for smooth vector fields, kernel Koopman operator approximations, and spectral convergence guarantees. These results underpin data-driven reduced-order modeling and forecasting for complex nonlinear dynamical systems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper adds representer theorems and approximation rates for operator-valued RKHS on time-varying fields, but the Koopman spectral claims look shaky without extra conditions on the time dependence.

read the letter

The main takeaway is that this work gives new representer theorems for time-aligned interpolation in operator-valued RKHS and pairs them with Sobolev-based approximation bounds plus kernel Koopman approximations. Those pieces are the actual additions, and they sit on top of existing RKHS and Koopman literature rather than replacing it. The derivations for the representer results and the rate bounds read as careful and internally consistent, which is the part that holds up best. They also spell out how the framework could support reduced-order modeling for nonlinear spatio-temporal systems, and the setup is clean enough that someone could try to implement the interpolation step directly. The credit here goes to the explicit handling of the time-alignment in the kernel and the attempt to quantify the approximation error under Sobolev regularity. That is useful bookkeeping even if the downstream applications stay theoretical for now. The softer part is the spectral convergence for the Koopman operator when the vector field depends explicitly on time. Standard arguments for spectral gaps or compactness in RKHS-Koopman settings rely on autonomous flows, and the non-autonomous cocycle case usually needs periodicity, slow variation, or bounds on the time derivative to close the estimates. The paper invokes Sobolev regularity but does not appear to derive or impose those extra controls, so the convergence guarantees may only hold in the autonomous limit or under unstated restrictions. That does not sink the whole paper, but it does mean the strongest claims need tighter hypotheses or a separate argument. This is for people already working on kernel methods for dynamical systems or data-driven reduced-order models. A reader who wants to see how OV RKHS can be adapted to time-varying interpolation will get concrete theorems to check; someone looking for turnkey forecasting tools will find less. The formal results are developed enough that a serious editor should send it to referees rather than desk-reject it, mainly to get the proofs and the time-dependence issue examined in detail.

Referee Report

1 major / 1 minor

Summary. The manuscript develops a framework for spatio-temporal prediction of time-varying vector fields using operator-valued reproducing kernel Hilbert spaces (OV RKHS). By integrating Sobolev regularity with Koopman operator theory, it establishes representer theorems for time-aligned OV RKHS interpolation, Sobolev approximation bounds, kernel Koopman operator approximations, and spectral convergence results to support data-driven reduced-order modeling and forecasting of nonlinear dynamical systems.

Significance. If the derivations hold, the work could provide a useful theoretical bridge between OV RKHS methods and Koopman spectral analysis for non-autonomous systems, offering approximation rates and convergence guarantees that may aid forecasting in applications such as fluid dynamics or control. The explicit use of Sobolev regularity to obtain representer theorems and spectral results is a potential strength if the time-dependent case is handled rigorously.

major comments (1)

[§4 (Kernel Koopman Operator Approximations)] §4 (Kernel Koopman Operator Approximations) and the spectral convergence theorem: the extension of standard Koopman spectral theory (typically formulated for autonomous flows) to explicitly time-dependent vector fields f(x,t) requires additional structure such as periodicity, slow variation, or uniform bounds on ∂_t f to obtain compactness or spectral gap estimates for the time-dependent cocycle. The Sobolev regularity assumption alone does not appear to supply these; the manuscript should state explicitly whether such conditions are imposed or derived, as this is load-bearing for the claimed spectral convergence guarantees.

minor comments (1)

[Preliminaries] The notation for the time-aligned OV RKHS and the precise definition of the operator-valued kernel incorporating both space and time should be stated more explicitly in the preliminaries to avoid ambiguity when applying the representer theorem.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful reading and constructive comments on our manuscript. We address the single major comment below and will make the requested clarifications in the revised version.

read point-by-point responses

Referee: [§4 (Kernel Koopman Operator Approximations)] §4 (Kernel Koopman Operator Approximations) and the spectral convergence theorem: the extension of standard Koopman spectral theory (typically formulated for autonomous flows) to explicitly time-dependent vector fields f(x,t) requires additional structure such as periodicity, slow variation, or uniform bounds on ∂_t f to obtain compactness or spectral gap estimates for the time-dependent cocycle. The Sobolev regularity assumption alone does not appear to supply these; the manuscript should state explicitly whether such conditions are imposed or derived, as this is load-bearing for the claimed spectral convergence guarantees.

Authors: We appreciate the referee highlighting this key technical point for non-autonomous systems. In the manuscript the time-dependent Koopman operator is defined as the cocycle generated by the flow of the explicitly time-varying vector field f(x,t). The Sobolev regularity assumptions on the vector fields, together with the compactness of the OV RKHS embedding into C^0, are used to obtain uniform bounds on the time variation of the approximated operators; these bounds are derived from the approximation rates proved in Section 3 rather than imposed separately. Nevertheless, to address the referee’s concern directly we will add an explicit remark (and a short lemma) in the revised Section 4 that states the derived uniform bounds on ∂_t f and explains how they guarantee the required compactness and spectral-gap estimates for the cocycle. This makes the load-bearing conditions fully transparent without altering the main theorems. revision: yes

Circularity Check

0 steps flagged

No circularity: theoretical integration remains self-contained

full rationale

The manuscript develops representer theorems, approximation rates, and spectral convergence by integrating Sobolev regularity assumptions with standard Koopman operator theory for dynamical systems. No equations or claims reduce a derived result to a fitted parameter, self-referential definition, or load-bearing self-citation whose validity depends on the present work. The abstract and claimed contributions rely on external mathematical structures (Sobolev spaces, RKHS interpolation, Koopman semigroups) whose independence is not contradicted by the provided text. The derivation chain therefore does not collapse to its inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The abstract references integration of Sobolev regularity and standard kernel and Koopman concepts but provides no explicit free parameters, invented entities, or detailed axioms; assessment is limited by absence of full text.

axioms (1)

domain assumption Time-varying vector fields possess Sobolev regularity sufficient for the OV RKHS framework.
Explicitly mentioned as integrated in the abstract to establish approximation bounds.

pith-pipeline@v0.9.0 · 5612 in / 1172 out tokens · 32243 ms · 2026-05-19T07:00:52.043820+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/AlexanderDuality.lean alexander_duality_circle_linking unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Theorem 4.2 (Spectral Convergence). ... ∥KN − K∥op → 0, and the eigenvalues {λ(N)k} and eigenfunctions {ϕ(N)k} of KN converge to those of K

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages

[1]

M. A. ´Alvarez, L. Rosasco, and N. D. Lawrence, Kernels for Vector-Valued Functions: A Review, Found. Trends Mach. Learn., vol. 4, no. 3, pp. 195–266, 2012. doi:10.1561/2200000036

work page doi:10.1561/2200000036 2012
[2]

Budiˇ si´ c, R

M. Budiˇ si´ c, R. Mohr, and I. Mezi´ c, Applied Koopmanism,Chaos: An Interdisciplinary Journal of Nonlinear Science, vol. 22, no. 4, p. 047510, 2012. doi:10.1063/1.4772195

work page doi:10.1063/1.4772195 2012
[3]

Carmeli, E

C. Carmeli, E. De Vito, and A. Toigo, Vector Valued Reproducing Kernel Hilbert Spaces and Universality , Analysis and Applications , vol. 8, no. 1, pp. 19–61, 2010. doi:10.1142/S0219530510001524

work page doi:10.1142/s0219530510001524 2010
[4]

Cucker and D.-X

F. Cucker and D.-X. Zhou, Learning Theory: An Approximation Theory Viewpoint . Cambridge: Cambridge University Press, 2007. doi:10.1017/CBO9780511721484

work page doi:10.1017/cbo9780511721484 2007
[5]

A. Das, A. Nair, C. Reddy, and S. L. Brunton, Learning Koopman Invariant Subspaces for Dynamic Mode Decomposition, IEEE Transactions on Automatic Control , vol. 65, no. 9, pp. 3859–3870, 2020. doi:10.1109/TAC.2020.2974508

work page doi:10.1109/tac.2020.2974508 2020
[6]

Das and A

S. Das and A. Mesbah, Optimal Control Using Koopman Operator and Model Predictive Control, IEEE Transactions on Automatic Control , vol. 66, no. 4, pp. 1647–1662, 2021. doi:10.1109/TAC.2020.3014252

work page doi:10.1109/tac.2020.3014252 2021
[7]

Fukumizu, F

K. Fukumizu, F. R. Bach, and M. I. Jordan, Kernel Dimension Reduction in Regression, Annals of Statistics, vol. 37, no. 4, pp. 1871–1905, 2009. doi:10.1214/08-AOS635

work page doi:10.1214/08-aos635 1905
[8]

Kadri, A

H. Kadri, A. Rakotomamonjy, F. Bach, and E. Pavez, Operator-Valued Kernels for Learning from Functional Response Data, J. Mach. Learn. Res., vol. 17, no. 20, pp. 1–54, 2016

work page 2016
[9]

Kato, Perturbation Theory for Linear Operators , 2nd ed

T. Kato, Perturbation Theory for Linear Operators , 2nd ed. Berlin: Springer-Verlag, 1995. ISBN: 978-3-540-58661-6 23

work page 1995
[10]

S. Klus, F. N¨ uske, S. Peitz, C. Sch¨ utte, and I. G. Kevrekidis, On Data-Driven Koopman Spectral Analysis for Dynamical Systems and Applications, Journal of Computational Dynamics , vol. 5, no. 1, pp. 1–36, 2018. doi:10.3934/jcd.2018001

work page doi:10.3934/jcd.2018001 2018
[11]

S. Klus, F. N¨ uske, S. Peitz, P. Koval, C. Sch¨ utte, and I. G. Kevrekidis, Kernel-Based Approximation of the Koopman Generator and the Perron-Frobenius Generator, Journal of Nonlinear Science, vol. 30, no. 1, pp. 347–386, 2020. doi:10.1007/s00332-019-09531-3

work page doi:10.1007/s00332-019-09531-3 2020
[12]

Koltchinskii, Random Matrix Approximation of Spectral Projection Operators: Ap- plications to Kernel PCA, Annals of Statistics , vol

V. Koltchinskii, Random Matrix Approximation of Spectral Projection Operators: Ap- plications to Kernel PCA, Annals of Statistics , vol. 28, no. 5, pp. 1694–1725, 2000. doi:10.1214/aos/1015957404

work page doi:10.1214/aos/1015957404 2000
[13]

Koltchinskii and M

V. Koltchinskii and M. Yuan, Risk Bounds for Regularized Least Squares Estimators , ESAIM: Probability and Statistics, vol. 4, pp. 47–66, 2000. doi:10.1051/ps:2000104

work page doi:10.1051/ps:2000104 2000
[14]

Korda and I

M. Korda and I. Mezi´ c, Convergence of Extended Dynamic Mode Decomposition to the Koopman Operator, SIAM Journal on Applied Dynamical Systems , vol. 17, no. 1, pp. 208–236,

work page
[15]

doi:10.1137/17M1135952

work page doi:10.1137/17m1135952
[16]

Korda and I

M. Korda and I. Mezi´ c, Linear Predictors for Nonlinear Dynamical Systems: Koop- man Operator Meets Model Predictive Control, Automatica, vol. 93, pp. 149–160, 2018. doi:10.1016/j.automatica.2018.04.041

work page doi:10.1016/j.automatica.2018.04.041 2018
[17]

W. Li, H. Owhadi, J. von Brecht, and A. Szlam, Sampling, Metric Entropy and Dimensionality Reduction, Applied and Computational Harmonic Analysis , vol. 43, no. 2, pp. 329–347, 2017. doi:10.1016/j.acha.2016.08.004

work page doi:10.1016/j.acha.2016.08.004 2017
[18]

X. Lian, X. Chen, Y. Yin, and X. Zhai, A Divide-and-Conquer Kernel Ridge Regression: Distributed Learning and Statistical Guarantees, Adv. Neural Inf. Process. Syst., vol. 34, pp. 29597–29609, 2021

work page 2021
[19]

Mezi´ c, Analysis of Fluid Flows via Spectral Properties of the Koopman Operator, Annual Review of Fluid Mechanics, vol

I. Mezi´ c, Analysis of Fluid Flows via Spectral Properties of the Koopman Operator, Annual Review of Fluid Mechanics, vol. 45, pp. 357–378, 2013. doi:10.1146/annurev-fluid-011212-140652

work page doi:10.1146/annurev-fluid-011212-140652 2013
[20]

C. A. Micchelli and M. Pontil, On Learning Vector-Valued Functions, Neural Computation, vol. 17, no. 1, pp. 177–204, 2005. doi:10.1162/0899766052530809

work page doi:10.1162/0899766052530809 2005
[21]

Rosasco, M

L. Rosasco, M. Belkin, and E. De Vito, On Learning with Integral Operators, Journal of Machine Learning Research, vol. 11, pp. 905–934, 2010

work page 2010
[22]

Rosasco, M

L. Rosasco, M. Belkin, E. De Vito, and A. Rudi, Vector Valued Learning with Operator-Valued Kernels, Foundations and Trends in Machine Learning , vol. 7, no. 3–4, pp. 177–266, 2016. doi:10.1561/2200000053

work page doi:10.1561/2200000053 2016
[23]

P. J. Schmid, Dynamic Mode Decomposition of Numerical and Experimental Data, Journal of Fluid Mechanics, vol. 656, pp. 5–28, 2010. doi:10.1017/S0022112010001217 24

work page doi:10.1017/s0022112010001217 2010
[24]

Steinwart and A

I. Steinwart and A. Christmann, Support Vector Machines , New York: Springer, 2008. doi:10.1007/978-0-387-77242-4

work page doi:10.1007/978-0-387-77242-4 2008
[25]

Wendland, Scattered Data Approximation, Cambridge University Press, Cambridge, UK,

H. Wendland, Scattered Data Approximation, Cambridge University Press, Cambridge, UK,

work page
[26]

doi:10.1017/CBO9780511617539

ISBN: 978-0-521-84695-9. doi:10.1017/CBO9780511617539

work page doi:10.1017/cbo9780511617539
[27]

M. O. Williams, I. G. Kevrekidis, and C. W. Rowley, A Data–Driven Approximation of the Koopman Operator: Extending Dynamic Mode Decomposition, in Proceedings of the 53rd IEEE Conference on Decision and Control , pp. 7199–7204, 2014. doi:10.1109/CDC.2014.7040178

work page doi:10.1109/cdc.2014.7040178 2014
[28]

M. O. Williams, I. G. Kevrekidis, and C. W. Rowley, A Data-Driven Approximation of the Koopman Operator: Extending Dynamic Mode Decomposition, Journal of Nonlinear Science , vol. 25, no. 6, pp. 1307–1346, 2015. doi:10.1007/s00332-015-9258-5 25

work page doi:10.1007/s00332-015-9258-5 2015

[1] [1]

M. A. ´Alvarez, L. Rosasco, and N. D. Lawrence, Kernels for Vector-Valued Functions: A Review, Found. Trends Mach. Learn., vol. 4, no. 3, pp. 195–266, 2012. doi:10.1561/2200000036

work page doi:10.1561/2200000036 2012

[2] [2]

Budiˇ si´ c, R

M. Budiˇ si´ c, R. Mohr, and I. Mezi´ c, Applied Koopmanism,Chaos: An Interdisciplinary Journal of Nonlinear Science, vol. 22, no. 4, p. 047510, 2012. doi:10.1063/1.4772195

work page doi:10.1063/1.4772195 2012

[3] [3]

Carmeli, E

C. Carmeli, E. De Vito, and A. Toigo, Vector Valued Reproducing Kernel Hilbert Spaces and Universality , Analysis and Applications , vol. 8, no. 1, pp. 19–61, 2010. doi:10.1142/S0219530510001524

work page doi:10.1142/s0219530510001524 2010

[4] [4]

Cucker and D.-X

F. Cucker and D.-X. Zhou, Learning Theory: An Approximation Theory Viewpoint . Cambridge: Cambridge University Press, 2007. doi:10.1017/CBO9780511721484

work page doi:10.1017/cbo9780511721484 2007

[5] [5]

A. Das, A. Nair, C. Reddy, and S. L. Brunton, Learning Koopman Invariant Subspaces for Dynamic Mode Decomposition, IEEE Transactions on Automatic Control , vol. 65, no. 9, pp. 3859–3870, 2020. doi:10.1109/TAC.2020.2974508

work page doi:10.1109/tac.2020.2974508 2020

[6] [6]

Das and A

S. Das and A. Mesbah, Optimal Control Using Koopman Operator and Model Predictive Control, IEEE Transactions on Automatic Control , vol. 66, no. 4, pp. 1647–1662, 2021. doi:10.1109/TAC.2020.3014252

work page doi:10.1109/tac.2020.3014252 2021

[7] [7]

Fukumizu, F

K. Fukumizu, F. R. Bach, and M. I. Jordan, Kernel Dimension Reduction in Regression, Annals of Statistics, vol. 37, no. 4, pp. 1871–1905, 2009. doi:10.1214/08-AOS635

work page doi:10.1214/08-aos635 1905

[8] [8]

Kadri, A

H. Kadri, A. Rakotomamonjy, F. Bach, and E. Pavez, Operator-Valued Kernels for Learning from Functional Response Data, J. Mach. Learn. Res., vol. 17, no. 20, pp. 1–54, 2016

work page 2016

[9] [9]

Kato, Perturbation Theory for Linear Operators , 2nd ed

T. Kato, Perturbation Theory for Linear Operators , 2nd ed. Berlin: Springer-Verlag, 1995. ISBN: 978-3-540-58661-6 23

work page 1995

[10] [10]

S. Klus, F. N¨ uske, S. Peitz, C. Sch¨ utte, and I. G. Kevrekidis, On Data-Driven Koopman Spectral Analysis for Dynamical Systems and Applications, Journal of Computational Dynamics , vol. 5, no. 1, pp. 1–36, 2018. doi:10.3934/jcd.2018001

work page doi:10.3934/jcd.2018001 2018

[11] [11]

S. Klus, F. N¨ uske, S. Peitz, P. Koval, C. Sch¨ utte, and I. G. Kevrekidis, Kernel-Based Approximation of the Koopman Generator and the Perron-Frobenius Generator, Journal of Nonlinear Science, vol. 30, no. 1, pp. 347–386, 2020. doi:10.1007/s00332-019-09531-3

work page doi:10.1007/s00332-019-09531-3 2020

[12] [12]

Koltchinskii, Random Matrix Approximation of Spectral Projection Operators: Ap- plications to Kernel PCA, Annals of Statistics , vol

V. Koltchinskii, Random Matrix Approximation of Spectral Projection Operators: Ap- plications to Kernel PCA, Annals of Statistics , vol. 28, no. 5, pp. 1694–1725, 2000. doi:10.1214/aos/1015957404

work page doi:10.1214/aos/1015957404 2000

[13] [13]

Koltchinskii and M

V. Koltchinskii and M. Yuan, Risk Bounds for Regularized Least Squares Estimators , ESAIM: Probability and Statistics, vol. 4, pp. 47–66, 2000. doi:10.1051/ps:2000104

work page doi:10.1051/ps:2000104 2000

[14] [14]

Korda and I

M. Korda and I. Mezi´ c, Convergence of Extended Dynamic Mode Decomposition to the Koopman Operator, SIAM Journal on Applied Dynamical Systems , vol. 17, no. 1, pp. 208–236,

work page

[15] [15]

doi:10.1137/17M1135952

work page doi:10.1137/17m1135952

[16] [16]

Korda and I

M. Korda and I. Mezi´ c, Linear Predictors for Nonlinear Dynamical Systems: Koop- man Operator Meets Model Predictive Control, Automatica, vol. 93, pp. 149–160, 2018. doi:10.1016/j.automatica.2018.04.041

work page doi:10.1016/j.automatica.2018.04.041 2018

[17] [17]

W. Li, H. Owhadi, J. von Brecht, and A. Szlam, Sampling, Metric Entropy and Dimensionality Reduction, Applied and Computational Harmonic Analysis , vol. 43, no. 2, pp. 329–347, 2017. doi:10.1016/j.acha.2016.08.004

work page doi:10.1016/j.acha.2016.08.004 2017

[18] [18]

X. Lian, X. Chen, Y. Yin, and X. Zhai, A Divide-and-Conquer Kernel Ridge Regression: Distributed Learning and Statistical Guarantees, Adv. Neural Inf. Process. Syst., vol. 34, pp. 29597–29609, 2021

work page 2021

[19] [19]

Mezi´ c, Analysis of Fluid Flows via Spectral Properties of the Koopman Operator, Annual Review of Fluid Mechanics, vol

I. Mezi´ c, Analysis of Fluid Flows via Spectral Properties of the Koopman Operator, Annual Review of Fluid Mechanics, vol. 45, pp. 357–378, 2013. doi:10.1146/annurev-fluid-011212-140652

work page doi:10.1146/annurev-fluid-011212-140652 2013

[20] [20]

C. A. Micchelli and M. Pontil, On Learning Vector-Valued Functions, Neural Computation, vol. 17, no. 1, pp. 177–204, 2005. doi:10.1162/0899766052530809

work page doi:10.1162/0899766052530809 2005

[21] [21]

Rosasco, M

L. Rosasco, M. Belkin, and E. De Vito, On Learning with Integral Operators, Journal of Machine Learning Research, vol. 11, pp. 905–934, 2010

work page 2010

[22] [22]

Rosasco, M

L. Rosasco, M. Belkin, E. De Vito, and A. Rudi, Vector Valued Learning with Operator-Valued Kernels, Foundations and Trends in Machine Learning , vol. 7, no. 3–4, pp. 177–266, 2016. doi:10.1561/2200000053

work page doi:10.1561/2200000053 2016

[23] [23]

P. J. Schmid, Dynamic Mode Decomposition of Numerical and Experimental Data, Journal of Fluid Mechanics, vol. 656, pp. 5–28, 2010. doi:10.1017/S0022112010001217 24

work page doi:10.1017/s0022112010001217 2010

[24] [24]

Steinwart and A

I. Steinwart and A. Christmann, Support Vector Machines , New York: Springer, 2008. doi:10.1007/978-0-387-77242-4

work page doi:10.1007/978-0-387-77242-4 2008

[25] [25]

Wendland, Scattered Data Approximation, Cambridge University Press, Cambridge, UK,

H. Wendland, Scattered Data Approximation, Cambridge University Press, Cambridge, UK,

work page

[26] [26]

doi:10.1017/CBO9780511617539

ISBN: 978-0-521-84695-9. doi:10.1017/CBO9780511617539

work page doi:10.1017/cbo9780511617539

[27] [27]

M. O. Williams, I. G. Kevrekidis, and C. W. Rowley, A Data–Driven Approximation of the Koopman Operator: Extending Dynamic Mode Decomposition, in Proceedings of the 53rd IEEE Conference on Decision and Control , pp. 7199–7204, 2014. doi:10.1109/CDC.2014.7040178

work page doi:10.1109/cdc.2014.7040178 2014

[28] [28]

M. O. Williams, I. G. Kevrekidis, and C. W. Rowley, A Data-Driven Approximation of the Koopman Operator: Extending Dynamic Mode Decomposition, Journal of Nonlinear Science , vol. 25, no. 6, pp. 1307–1346, 2015. doi:10.1007/s00332-015-9258-5 25

work page doi:10.1007/s00332-015-9258-5 2015