Pith. sign in

REVIEW 2 major objections 5 minor 26 references

Prediction accuracy does not certify learning stability

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · glm-5.2

2026-07-08 04:58 UTC pith:VWDIGIZ5

load-bearing objection Clean deterministic ISS framework for Koopman learning control; finite-sample radii are conservative but honestly scoped. the 2 major comments →

arxiv 2607.06459 v1 pith:VWDIGIZ5 submitted 2026-07-07 eess.SY cs.SY

Input-to-State Stability Certification via Projection Residuals for Koopman Learning Control of Nonlinear Repetitive Systems

Yue Wu , Ye Cao , Jianfu Cao This is my paper
classification eess.SY cs.SY
keywords projectionchannelkoopmanlearningresidualsselectederrornonlinear
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper proves that a Koopman model with small held-out prediction residuals can still fail to produce a stable learning controller, because stability along the trial axis requires a positive margin on the selected finite-horizon input-output channel and a bounded projection residual measuring the gap between requested output corrections and what the constrained actuators can actually deliver. The authors formulate the stacked tracking error as the state of a discrete-time system indexed by trial number, treat Koopman residuals, reset mismatch, channel uncertainty, projection residuals, deployment shifts, and numerical tolerances as disturbance inputs, and prove that the learning error decays geometrically to a computable ultimate band whose radius is the sum of all these perturbation budgets divided by one minus the learning gain. A finite-sample implementation uses split-conformal calibration on frozen controllers to construct the residual budget with probabilistic coverage, and a rejection-capable certification protocol discards any candidate whose channel margin, projection closure, or certified band fails to meet the requirements.

Core claim

The central object is the projection residual, defined as the distance from the requested output increment to the constrained reachable set of the learned finite-horizon channel. This residual is a structural actuation limit, not a numerical error, and it enters the ISS budget as an irreducible contribution to the ultimate error band. The paper proves that if the learned channel's minimum singular value minus its certified perturbation radius is nonpositive, no stability certificate can be issued regardless of how small the prediction residuals are, formally separating model fit from learning certifiability.

What carries the argument

The practical ISS bound (Theorem 1, Eq. 22): ||E_bar_k|| <= lambda^k ||E_bar_0|| + ((1-lambda^k)/(1-lambda)) * w_bar, where w_bar = xi_0 + p_bar + eps_G * U_bar_Delta + eps_O * Z_bar_0. The learning gain lambda controls geometric decay rate; w_bar aggregates residual, projection, channel, and reset budgets into the ultimate band radius w_bar/(1-lambda).

Load-bearing premise

The certified deployment event (Assumption 1) requires that the true channel deviates from the learned channel by no more than a certified spectral-norm radius, and that the learned channel's minimum singular value exceeds this radius by a positive margin. If the true channel deviates more than the certified radius, the entire stability budget is invalid.

What would settle it

Construct a system where the Koopman predictor has arbitrarily small held-out prediction residuals but the selected finite-horizon channel has a near-zero minimum singular value relative to its perturbation radius. The theory predicts this predictor must be rejected as non-certifiable despite its excellent prediction accuracy.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • A learned Koopman predictor with excellent prediction accuracy can be rejected from deployment if its selected finite-horizon channel is weak, meaning the actuators cannot reliably produce the output corrections the learning law requests.
  • Each certification failure can be traced to a specific cause: large residual score implicates the predictor or noise level; nonpositive channel margin implicates data informativeness; large projection residual implicates actuator constraints; dominant shift or numerical terms implicate the implementation protocol.
  • The certified ultimate band is not zero under realistic conditions because finite samples, calibration scores, and deployment disturbances are nonzero, making this a practical rather than idealized stability statement.
  • The certification protocol is rejection-capable: it returns 'Not Certified' whenever the contraction condition, channel margin, request closure, or certified band requirement fails, rather than deploying an uncertified controller.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. This paper develops an input-to-state stability (ISS) certificate for Koopman-based learning control of unknown nonlinear repetitive systems operating over finite trial horizons. The central idea is to treat the selected stacked tracking error along the learning-trial axis as the state of a discrete-time system, with Koopman residuals, projection residuals, channel uncertainty, reset mismatch, deployment shifts, and numerical tolerances acting as ISS inputs. The deterministic result (Theorem 1) provides a parameter-free ultimate band derived from independently certifiable quantities. A finite-sample extension (Theorem 2) uses split conformal prediction at the episode level to calibrate the residual budget under exchangeability assumptions. The paper also identifies projection residuals and selected-channel margins as necessary certificate objects, showing that small prediction residuals alone are insufficient for learning stability certification. Numerical experiments on Duffing-type and other nonlinear repetitive systems audit the certificate components.

Significance. The paper makes a clear conceptual contribution by separating model fit from learning certifiability in Koopman-based control. The deterministic ISS certificate (Theorem 1, Eq. 22) is a clean, parameter-free derivation: the budget w_bar is constructed from independently certifiable quantities (residuals, channel margins, projection distances). The projection-residual analysis (Propositions 1-2) correctly identifies a structural obstruction that prediction loss alone cannot capture. The finite-sample extension (Theorem 2) correctly applies split conformal prediction at the episode level under Assumption 2, with the controller frozen before calibration. The paper ships reproducible code and data, and the numerical experiments are structured as certificate audits rather than performance benchmarks, which is appropriate for the claims. The weak-channel rejection mechanism (Proposition 3) is a useful and falsifiable diagnostic.

major comments (2)
  1. Lemma 3, Eqs. (27)-(30): The perturbation radii D_A(j) accumulate terms involving ||Â||^(j-1-ℓ) * (||Â|| + ε_A)^ℓ, which grows exponentially with horizon H when ||Â|| > 1. Since Koopman lifted matrices frequently have spectral radii exceeding 1 (the physical plant may be stable while the lifted representation is not), the stacked radii ε_{G,H,β} and ε_{O,H,β} in Eq. (30) can become extremely large for moderate H. The numerical experiments (Table 4) report w_bar values around 0.4-0.6 with bands around 0.8-1.1, which are non-trivial but relatively wide. The paper should explicitly discuss the regime where these radii yield non-vacuous certificates — for instance, characterizing the relationship between H, ||Â||, and the resulting band size, or noting whether the experiments operate in a regime where ||Â|| < 1. This is load-bearing because Theorem 2's practical value depends on these radii.
  2. Table 6: Several rows report channel and reset terms as exactly 0.000 (e.g., all disturbance and noise sweep rows). Given that Lemma 3's radii are generically nonzero for any finite-data identified model, these zero values require clarification. Are these set to zero because the channel perturbation is negligible relative to the reported precision, or are they excluded from the budget for another reason? If the former, the table should indicate this; if the latter, the relationship between the finite-sample budget (Eq. 33) and the reported numbers needs clarification.
minor comments (5)
  1. The paper lists no ad-hoc axioms or invented entities in its formal ledger, which is commendable. However, the relationship between Assumption 0 (Certified nonlinear repetitive task class) and Assumption 1 (Certified deployment event) could be stated more precisely: Assumption 0 defines the task class, while Assumption 1 defines the per-trial event under which the certificate holds. A brief sentence clarifying that Assumption 1 is verified through the finite-sample machinery of Theorem 2 would help the reader.
  2. Eq. (33) and Remark 3 use slightly different orderings of the budget terms (Δw and Δ_{num,β} appear in different positions). Consistent ordering would improve readability.
  3. Table 2 lists four systems but Tables 4-9 primarily report results for the Duffing system and its variants. A brief note on which table corresponds to which system in Table 2 would help cross-referencing.
  4. The CRediT statement and AI declaration are appropriate and transparent.
  5. Minor typographic issue: 'certifia-bility' in the Highlights section (line break artifact).

Simulated Author's Rebuttal

2 responses · 0 unresolved

The referee recommends minor revision and identifies two major comments: (1) the exponential growth of perturbation radii in Lemma 3 when the lifted matrix norm exceeds 1, and (2) the zero-valued channel and reset terms in Table 6. Both comments are well-taken. We will add an explicit discussion of the non-vacuous certification regime and clarify the table reporting conventions.

read point-by-point responses

  1. Referee: Lemma 3, Eqs. (27)-(30): The perturbation radii D_A(j) accumulate terms involving ||Â||^(j-1-ℓ) * (||Â|| + ε_A)^ℓ, which grows exponentially with horizon H when ||Â|| > 1. Since Koopman lifted matrices frequently have spectral radii exceeding 1 (the physical plant may be stable while the lifted representation is not), the stacked radii ε_{G,H,β} and ε_{O,H,β} in Eq. (30) can become extremely large for moderate H. The numerical experiments (Table 4) report w_bar values around 0.4-0.6 with bands around 0.8-1.1, which are non-trivial but relatively wide. The paper should explicitly discuss the regime where these radii yield non-vacuous certificates — for instance, characterizing the relationship between H, ||Â||, and the resulting band size, or noting whether the experiments operate in a regime where ||Â|| < 1. This is load-bearing because Theorem 2's practical value depends on these radii.

    Authors: The referee is correct that the perturbation radii in Lemma 3 grow exponentially with H when ||Â|| > 1, and that this is a genuine limitation of the bound rather than a presentation artifact. We will add an explicit discussion of this regime. Specifically, we plan to add a remark after Lemma 3 stating three points. First, the bound (30) is conservative because it uses induced-norm submultiplicativity rather than spectral-radius arguments; when the lifted matrix  has spectral radius below 1 but induced norm above 1, the bound may still grow with H even though the true stacked error does not. Second, in the numerical experiments, the identified lifted matrices for the Duffing, cubic damping, and nonlinear servo systems do have ||Â|| moderately above 1 (typically 1.1-1.5 in the induced 2-norm), but the horizons N = 70-90 are short enough that the stacked radii remain non-vacuous when combined with the other budget terms. Third, the certificate becomes vacuous when the product of ||Â||^H and the one-step error radii exceeds the channel margin, which is a fundamental limitation of any finite-data perturbation bound of this type. We will also note that the radii ε_{G,H,β} and ε_{O,H,β} can be replaced by externally audited bounds (as already mentioned in Lemma 3), which may be tighter than the submultiplicative construction when the lifted spectrum is favorable. We agree this discussion is load-bearing for Theorem 2 and will add it to the revised manuscript. revision: yes

  2. Referee: Table 6: Several rows report channel and reset terms as exactly 0.000 (e.g., all disturbance and noise sweep rows). Given that Lemma 3's radii are generically nonzero for any finite-data identified model, these zero values require clarification. Are these set to zero because the channel perturbation is negligible relative to the reported precision, or are they excluded from the budget for another reason? If the former, the table should indicate this; if the latter, the relationship between the finite-sample budget (Eq. 33) and the reported numbers needs clarification.

    Authors: The referee is correct to flag this. The zero values in the channel and reset columns of Table 6 arise because the sweep experiment was designed to isolate the effect of a single perturbation source (disturbance, noise, or mismatch) on the budget, and in those rows the channel and reset perturbation radii were not separately computed — they were set to zero to isolate the swept variable. This is a reporting choice that should have been stated explicitly. In the revised manuscript, we will add a note to Table 6 clarifying that the sweep rows hold the channel and reset terms at zero to isolate the effect of the swept variable, and that the full budget (including nonzero channel and reset terms) is reported separately in Table 4. We will also verify that the column headers are unambiguous about which terms from Eq. (33) are being reported in each table. If the referee's concern is that the zero values misrepresent the finite-sample budget as defined in Eq. (33), we agree that the current presentation is unclear and will fix it. revision: yes

Circularity Check

0 steps flagged

No circularity: the ISS certificate is a parameter-free derivation from independently certifiable quantities, and the finite-sample extension uses split conformal prediction on held-out calibration episodes.

full rationale

The paper's central deterministic result (Theorem 1, Eq. 22) is a straightforward, self-contained derivation: Lemma 2 substitutes the true channel (Eq. 6) into the error recursion, cancels nominal terms, and bounds the residual by the triangle inequality using Assumption 1's componentwise bounds (Eq. 16). The resulting ISS bound ||E_bar_k|| <= lambda^k ||E_bar_0|| + ((1-lambda^k)/(1-lambda)) * w_bar is a standard scalar comparison recursion iteration with w_bar = xi_0 + p_bar + eps_G * U_bar_Delta + eps_O * Z_bar_0 (Eq. 21). No term in w_bar is defined in terms of the ISS conclusion; each is an independently certifiable or calibrated quantity (residual bound, projection distance, channel perturbation radius, reset bound). The finite-sample extension (Theorem 2, Eq. 33) replaces deterministic bounds with split-conformal calibrated quantities on held-out episodes (D_cal) that are explicitly separated from the design split (D_des) and frozen before calibration (Algorithm 1, lines 15-21). The conformal score (Eq. 31) measures complete-episode mismatch under the frozen controller, not a fitted parameter repackaged as a prediction. Lemma 3's perturbation radii (Eqs. 27-30) propagate one-step identification errors via submultiplicative norms — conservative but not circular. Propositions 1-5 establish structural properties (projection obstruction, channel margin, request closure) that are independent mathematical facts, not self-citations. The numerical experiments (Tables 4-9) audit certificate components rather than fitting and re-predicting the same data. No self-citation chain is load-bearing for the central claim. The derivation is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

4 free parameters · 6 axioms · 0 invented entities

The paper introduces no new physical entities, particles, forces, or dimensions. The 'projection residual' p_k and the 'selected-channel margin' gamma_G are new certificate objects but they are computed from existing mathematical quantities (the learned channel G_hat and the admissible update set U_Delta), not postulated entities. The Koopman lifting map psi is standard in the Koopman literature.

free parameters (4)
  • lambda (learning gain)
    Constrained to ||Lambda||_inf <= lambda < 1; chosen from candidate set Q in Algorithm 1. Not fitted to data but selected by the designer.
  • mu (regularization)
    Regularization parameter in the constrained least-squares update (Eq. 8); selected from candidate set Q.
  • R (certified error radius)
    Used in Definition 2 for the request set and in Proposition 4 for invariance; a design parameter for the certified ball.
  • beta_cal, beta_par, beta_shift, beta_proj, beta_num (risk levels)
    Declared risk levels for the finite-sample certificate in Theorem 2; chosen by the practitioner.
axioms (6)
  • domain assumption Assumption 0: The task protocol is resettable and the reference trajectory is repeated; state, input, output, disturbance, reset mismatch, and actuator update are restricted to certified bounded sets.
    Section 2. Required for the finite-horizon formulation and for the ISS certificate to apply.
  • domain assumption Assumption 1: The learned model, true selected channel, and deployment protocol satisfy the certified deployment event (Eq. 16-17) with bounded residuals, channel perturbation, reset mismatch, and positive channel margin.
    Section 4. The load-bearing premise for Theorem 1; without it the ISS recursion does not close.
  • domain assumption Assumption 2: Conditional on the fitted Koopman model and frozen controller, calibration episodes and deployment episode are exchangeable under the declared resettable task protocol.
    Section 4. Required for the split conformal prediction in Theorem 2; stated at episode level, not sample level.
  • domain assumption The within-trial plant is not assumed to be globally ISS; the certificate concerns only the selected stacked tracking error along the learning-trial axis.
    Section 2, Assumption 0. Separates trial-axis stability from within-trial internal stability.
  • standard math Standard ISS theory for discrete-time nonlinear systems (Sontag, Jiang-Teel-Praly).
    Section 1, references [1-4]. Used as the mathematical framework for the trial-axis stability certificate.
  • standard math Split conformal prediction provides distribution-free coverage under exchangeability (Vovk et al., Lei et al.).
    Section 4, references [25-26]. Used in Theorem 2 for the finite-sample residual budget.

pith-pipeline@v1.1.0-glm · 18828 in / 3223 out tokens · 440552 ms · 2026-07-08T04:58:27.486589+00:00 · methodology

0 comments
read the original abstract

This paper studies input-to-state stability (ISS) certification for data-driven Koopman learning control of unknown discrete-time nonlinear repetitive systems over finite trial horizons. Rather than proposing a new learning law, we certify when a fixed Koopman-assisted constrained update yields practical stability of the selected tracking error along the trial axis. Prediction accuracy alone is insufficient for this purpose: the selected finite-horizon input-output channel must have a positive margin, and the unreachable component of the requested output increment must be accounted for through a projection residual. Thus, a Koopman predictor with small held-out prediction residuals may still fail the learning-stability certificate if its selected channel is weak. We formulate the selected stacked tracking error as the state of a discrete-time learning-axis system and treat Koopman residuals, reset mismatch, channel uncertainty, projection residuals, deployment shifts, and numerical tolerances as ISS inputs. The deterministic result gives a practical ISS estimate from the initial learning error to an explicit ultimate band. A finite-sample implementation constructs an episode-level residual bound under a fixed controller and combines it with reported channel, projection, shift, and numerical margins. Numerical checks on nonlinear repetitive systems support the predicted residual-to-band scaling, weak-channel rejection, projection closure, and ultimate-band coverage.

Figures

Figures reproduced from arXiv: 2607.06459 by Jianfu Cao, Ye Cao, Yue Wu.

Figure 1
Figure 1. Figure 1: Finite-sample ISS budget decomposition. Each bar is the sum of the episode [PITH_FULL_IMAGE:figures/full_fig_p020_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Episode-level split-conformal residual audit. The controller is frozen before [PITH_FULL_IMAGE:figures/full_fig_p021_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Closed-loop state, output, input, and input-update traces for the residual [PITH_FULL_IMAGE:figures/full_fig_p022_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Residual-to-band validation. The certified band grows with the aggregate per [PITH_FULL_IMAGE:figures/full_fig_p023_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Selected-channel certification ablation. The weak-channel case is rejected be [PITH_FULL_IMAGE:figures/full_fig_p024_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Candidate rejection cascade. The certificate rejects candidates at the first failed [PITH_FULL_IMAGE:figures/full_fig_p025_6.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

26 extracted references · 26 canonical work pages · 2 internal anchors

  1. [1]

    Smooth stabilization implies coprime factorization,

    E. D. Sontag, “Smooth stabilization implies coprime factorization,”IEEE Transactions on Automatic Control, vol. 34, no. 4, pp. 435–443, 1989, doi: 10.1109/9.28018

    work page doi:10.1109/9.28018 1989

  2. [2]

    On characterizations of the input-to-state stability property,

    E. D. Sontag and Y. Wang, “On characterizations of the input-to-state stability property,”Systems & Control Letters, vol. 24, no. 5, pp. 351–359, 1995, doi: 10.1016/0167-6911(94)00050-6

    work page doi:10.1016/0167-6911(94)00050-6 1995

  3. [3]

    Small-gain theorem for ISS systems and applications,

    Z.-P. Jiang, A. R. Teel, and L. Praly, “Small-gain theorem for ISS systems and applications,”Math- ematics of Control, Signals and Systems, vol. 7, no. 2, pp. 95–120, 1994, doi: 10.1007/BF01211469

    work page doi:10.1007/bf01211469 1994

  4. [4]

    Input-to-state stability for discrete-time nonlinear systems,

    Z.-P. Jiang and Y. Wang, “Input-to-state stability for discrete-time nonlinear systems,”Automatica, vol. 37, no. 6, pp. 857–869, 2001, doi: 10.1016/S0005-1098(01)00028-0

    work page doi:10.1016/s0005-1098(01)00028-0 2001

  5. [5]

    A characterization of integral input-to-state stability,

    D. Angeli, E. D. Sontag, and Y. Wang, “A characterization of integral input-to-state stability,”IEEE Transactions on Automatic Control, vol. 45, no. 6, pp. 1082–1097, 2000, doi: 10.1109/9.863594

    work page doi:10.1109/9.863594 2000

  6. [6]

    A Lyapunov approach to incremental stability properties,

    D. Angeli, “A Lyapunov approach to incremental stability properties,”IEEE Transactions on Auto- matic Control, vol. 47, no. 3, pp. 410–421, 2002, doi: 10.1109/9.989067

    work page doi:10.1109/9.989067 2002

  7. [7]

    Small gain theorems for large scale systems and construction of ISS Lyapunov functions,

    S. N. Dashkovskiy, B. S. Rüffer, and F. R. Wirth, “Small gain theorems for large scale systems and construction of ISS Lyapunov functions,”SIAM Journal on Control and Optimization, vol. 48, no. 6, pp. 4089–4118, 2010, doi: 10.1137/090746483

    work page doi:10.1137/090746483 2010

  8. [8]

    Characterizations of input-to-state stability for infinite-dimensional systems,

    A. Mironchenko and F. Wirth, “Characterizations of input-to-state stability for infinite-dimensional systems,”IEEE Transactions on Automatic Control, vol. 63, no. 6, pp. 1692–1707, 2018, doi: 10.1109/TAC.2017.2756341

    work page doi:10.1109/tac.2017.2756341 2018

  9. [9]

    Bettering operation of robots by learning,

    S. Arimoto, S. Kawamura, and F. Miyazaki, “Bettering operation of robots by learning,”Journal of Robotic Systems, vol. 1, no. 2, pp. 123–140, 1984, doi: 10.1002/rob.4620010203

    work page doi:10.1002/rob.4620010203 1984

  10. [10]

    K. L. Moore,Iterative Learning Control for Deterministic Systems. London, U.K.: Springer-Verlag, 1993, doi: 10.1007/978-1-4471-1912-8

    work page doi:10.1007/978-1-4471-1912-8 1993

  11. [11]

    Iterative learning control using optimal feedback and feedforward actions,

    N. Amann, D. H. Owens, and E. Rogers, “Iterative learning control using optimal feedback and feedforward actions,”International Journal of Control, vol. 65, no. 2, pp. 277–293, 1996, doi: 10.1080/00207179608921697

    work page doi:10.1080/00207179608921697 1996

  12. [12]

    Robust iterative learning control with applications to injection mold- ing process,

    F. Gao, Y. Yang, and C. Shao, “Robust iterative learning control with applications to injection mold- ing process,”Chemical Engineering Science, vol. 56, no. 24, pp. 7025–7034, 2001, doi: 10.1016/S0009- 2509(01)00339-6

    work page doi:10.1016/s0009- 2001

  13. [13]

    Feasibility studies for the measurement of time-like proton electromagnetic form factors from $\bar{p}p \rightarrow \mu^+\mu^-$ at $\overline{\textrm{P}}\textrm{ANDA}$ at FAIR

    D. A. Bristow, M. Tharayil, and A. G. Alleyne, “A survey of iterative learning control,”IEEE Control Systems Magazine, vol. 26, no. 3, pp. 96–114, 2006, doi: 10.1109/MCS.2006.1636313

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1109/mcs.2006.1636313 2006

  14. [14]

    Hamiltonian systems and transformation in Hilbert space,

    B. O. Koopman, “Hamiltonian systems and transformation in Hilbert space,”Proceedings of the National Academy of Sciences of the United States of America, vol. 17, no. 5, pp. 315–318, 1931, doi: 10.1073/pnas.17.5.315

    work page doi:10.1073/pnas.17.5.315 1931

  15. [15]

    Spectral analysis of nonlinear flows,

    C. W. Rowley, I. Mezić, S. Bagheri, P. Schlatter, and D. S. Henningson, “Spectral analysis of nonlinear flows,”Journal of Fluid Mechanics, vol. 641, pp. 115–127, 2009, doi: 10.1017/S0022112009992059

    work page doi:10.1017/s0022112009992059 2009

  16. [16]

    A Data-Driven Approximation of the Koopman Operator: Extending Dynamic Mode Decomposition

    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

    work page internal anchor Pith review doi:10.1007/s00332-015-9258-5 2015

  17. [17]

    Koopman invariant subspaces and finite linear representations of nonlinear dynamical systems for control,

    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, vol. 11, no. 2, e0150171, 2016, doi: 10.1371/journal.pone.0150171

    work page doi:10.1371/journal.pone.0150171 2016

  18. [18]

    Dynamic mode decomposition with control,

    J. L. Proctor, S. L. Brunton, and J. N. Kutz, “Dynamic mode decomposition with control,”SIAM Journal on Applied Dynamical Systems, vol. 15, no. 1, pp. 142–161, 2016, doi: 10.1137/15M1013857

    work page doi:10.1137/15m1013857 2016

  19. [19]

    Linear Predictors for Nonlinear Dynamical Systems: Koopman Operator Meets Model Predictive Control

    M. Korda and I. Mezić, “Linear predictors for nonlinear dynamical systems: Koopman operator meets model predictive control,”Automatica, vol. 93, pp. 149–160, 2018, doi: 10.1016/j.automatica.2018.03.046

    work page doi:10.1016/j.automatica.2018.03.046 2018

  20. [20]

    Koopman-based lifting techniques for nonlinear systems identi- fication,

    A. Mauroy and J. Goncalves, “Koopman-based lifting techniques for nonlinear systems identi- fication,”IEEE Transactions on Automatic Control, vol. 65, no. 6, pp. 2550–2565, 2020, doi: 10.1109/TAC.2019.2941433

    work page doi:10.1109/tac.2019.2941433 2020

  21. [21]

    Finite data error bounds for Koopman-based prediction and control,

    F. Nüske, S. Peitz, F. Philipp, M. Schaller, and K. Worthmann, “Finite data error bounds for Koopman-based prediction and control,”Journal of Nonlinear Science, vol. 33, no. 1, 2023, doi: 10.1007/s00332-022-09862-1

    work page doi:10.1007/s00332-022-09862-1 2023

  22. [22]

    On the sample complexity of the linear quadratic regulator,

    S. Dean, H. Mania, N. Matni, B. Recht, and S. Tu, “On the sample complexity of the linear quadratic regulator,”Foundations of Computational Mathematics, vol. 20, no. 4, pp. 633–679, 2020, doi: 10.1007/s10208-019-09426-y. 27

    work page doi:10.1007/s10208-019-09426-y 2020

  23. [23]

    Formulas for data-driven control: Stabilization, optimality, and ro- bustness,

    C. De Persis and P. Tesi, “Formulas for data-driven control: Stabilization, optimality, and ro- bustness,”IEEE Transactions on Automatic Control, vol. 65, no. 3, pp. 909–924, 2020, doi: 10.1109/TAC.2019.2959924

    work page doi:10.1109/tac.2019.2959924 2020

  24. [24]

    Data-driven model predictive control with stability and robustness guarantees,

    J. Berberich, J. Köhler, M. A. Müller, and F. Allgöwer, “Data-driven model predictive control with stability and robustness guarantees,”IEEE Transactions on Automatic Control, vol. 66, no. 4, pp. 1702–1717, 2021, doi: 10.1109/TAC.2020.3000182

    work page doi:10.1109/tac.2020.3000182 2021

  25. [25]

    V. Vovk, A. Gammerman, and G. Shafer,Algorithmic Learning in a Random World. New York, NY, USA: Springer, 2005, doi: 10.1007/b106715

    work page doi:10.1007/b106715 2005

  26. [26]

    Journal of the American Statistical Association , author =

    J.Lei, M.G’Sell, A.Rinaldo, R.Tibshirani, andL.Wasserman, “Distribution-freepredictiveinference for regression,”Journal of the American Statistical Association, vol. 113, no. 523, pp. 1094–1111, 2018, doi: 10.1080/01621459.2017.1307116. 28

    work page doi:10.1080/01621459.2017.1307116 2018