Pith. sign in

REVIEW 4 major objections 7 minor 1 cited by

Estimating Varying Parameters in Dynamical Systems: A Modular Framework Using Switch Detection, Optimization, and Sparse Regression

T0 review · 4 major / 7 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A modular framework combining switch detection, optimization, and sparse regression can estimate time-varying parameters and their switch times in known dynamical systems.

desk verdict A useful modular pipeline for estimating time-varying parameters, but internal inconsistencies and an untested unknown-switch-count case keep it from being fully convincing. read the letter →

arxiv 2412.16198 v3 pith:JQKO2OYP submitted 2024-12-16 math.OC math.DS

Jamiree Harrison , Enoch Yeung This is my paper · ORCID
classification math.OCmath.DS MSC 93B3062M1090C5665L09
keywords parameterestimationparameter-varyingsystemsswitchdetectionchangepointbinarysegmentationsparseregressiondynamicalpiecewiseconstantparameters
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish a practical, modular answer to a hard inverse problem: given data from a dynamical system whose equations are known but whose parameters change over time or space, can we recover both the parameter values and the times at which they switch? The authors argue that for the class of systems with piecewise-constant switching parameters, the answer is yes, using a pipeline of switch detection, numerical integration, and gradient-free optimization. They extend the same pipeline to continuously varying parameters by sampling the parameter finely and fitting dictionary functions with sparse regression. The payoff is a general estimation tool that works across ODEs and PDEs, including a genetic toggle switch previously shown to be structurally unidentifiable, with demonstrated accuracy on switch locations, parameter values, and reconstructed trajectories.

What carries the argument

The central object is the modular pipeline itself: switch detection, implemented with binary segmentation (binseg) on each state trajectory using a noise-level hyperparameter $\sigma$ and a switch-gap $s_g$; data segmentation at detected switch locations; per-segment parameter estimation by minimizing the squared trajectory error $\|X_{\mathrm{data}} - X_{\mathrm{model}}\|_2^2$, where $X_{\mathrm{model}}$ comes from numerically integrating the known dynamics; and, for continuous parameters, sparse regression that fits the sampled parameters as a sparse linear combination of dictionary functions $\mathcal{D}_\theta w$. The framework's modularity is itself load-bearing: users can swap the switch detection, integration, or optimization method, but the overall estimate is only as strong as the weakest sub-step.

What would settle it

Run the protein synthesis model of equations (19)–(22) with white noise of standard deviation $\sigma = 2.25$, leaving the number of switches unknown and letting the switch-detection step estimate it from the data. If the detected switch location moves away from the true switch at $t = 12$ by more than the switch gap setting, the estimated synthesis rates $\alpha_m$ and $\alpha_p$ on the first interval will shift measurably from their true value of 4.0, contradicting the claim of accurate switch-and-parameter estimation in the noisy case.

Watch

Extended reading notes

Core claim

The paper's central claim is that a modular framework—switch detection by binary segmentation, per-segment numerical integration, and optimization-based parameter estimation—can accurately recover both the values and the switch locations of time-varying parameters in dynamical systems with known model structure. For continuously varying parameters, the framework samples the parameter over fine segments and applies dictionary-based sparse regression with trigonometric and polynomial functions to recover a continuous parameter function. The authors validate this on a parameter-varying genetic toggle switch, a protein synthesis model with uniform and non-uniform switches, a heat equation with a time-varying diffusion coefficient, and an advection-diffusion equation with a continuously varying advection coefficient, reporting exact switch detection on the noiseless examples. They also quantify the effect of measurement noise, showing a cascading propagation of error from switch detection to parameter fitting to trajectory reconstruction.

Load-bearing premise

The framework's success depends on the switch-detection step finding the correct number and locations of switches in the state data; if noise causes it to miss or over-detect switches, the subsequent parameter estimates are fit to the wrong intervals and the whole estimate collapses.

Editorial extensions

If this is right

  • Users with known model structure and switching parameters can recover parameter values and switch times directly from snapshot data, without manual segmentation.
  • The heat-equation example shows the framework applies to PDEs with time-varying coefficients when spatial derivatives are discretized by the method of lines.
  • The framework retains accuracy on a toggle-switch model that is structurally unidentifiable, suggesting that lack of identifiability need not prevent practical parameter estimation.
  • Because the framework is modular, any improvement in change-point detection, numerical integration, or optimization automatically improves the end-to-end estimation.
  • The documented noise sensitivity means users should expect error to grow with noise in a predictable cascade: wrong switches lead to wrong segments, which lead to wrong parameters and trajectories.

Reading between the lines

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

  • One extension we see: the switch-detection hyperparameters $\sigma$ and $s_g$ could be tuned automatically by cross-validating the final trajectory error on held-out data, removing the current reliance on user-chosen noise levels.
  • The cascading failure mode the paper identifies suggests that a robust variant should use redundant switch detection—for example, consensus across multiple change-point algorithms or across state trajectories—before parameter fitting.
  • We would expect the sparse-regression extension to generalize to spatially or temporally varying PDE coefficients beyond the advection-diffusion example, since the method only requires a well-chosen dictionary of candidate functions.
  • The paper notes the equivalence between varying parameters and inputs; a natural testable extension is to apply the same pipeline to estimate time-varying external inputs or controls in feedback systems.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper proposes a modular framework for estimating time-varying parameters in dynamical systems whose model structure is known. In the primary setting the parameters are piecewise constant, and the framework detects discrete switch locations from the state trajectories using binary segmentation, segments the data accordingly, and fits parameters on each interval by numerical integration plus derivative-free optimization (Nelder-Mead, Powell, or Differential Evolution). For continuously varying parameters, the framework segments the data into N/6 intervals, obtains per-interval parameter samples, and applies sparse dictionary regression. The method is demonstrated on ODE examples (a parameter-varying toggle switch, a gene-expression model with non-uniform switches) and PDE examples (the heat equation and the advection-diffusion equation), together with a measurement-noise study. Performance is reported through four metrics: parameter error Ep, trajectory error Et, switch-number error ENs, and Hausdorff distance Hs. The central claim is that the framework accurately estimates switching parameters, their switch locations, and continuously varying parameter functions for a broad problem class.

Significance. If the reported accuracies held over the stated problem class, the framework would be a practically useful contribution because it combines existing building blocks in a modular way, covers ODEs and PDEs, and is accompanied by code. The noiseless synthetic demonstrations that achieve Hs=0 and ENs=0 on several examples are encouraging, and the paper is explicit about some limitations, notably the cascading error propagation summarized in Eq. (23). However, the load-bearing question of switch-number estimation when N_s is unknown is not actually tested: the noisy-data experiments fix N_s=1, several reported configurations are internally inconsistent (Sections 3.3 and 3.4), and the phase reconstruction in the sparse-regression example is incorrect. These issues mean that the current evidence supports only a narrower claim, namely accurate estimation when the user supplies the switch count or successfully tunes the switch-detection hyperparameters.

major comments (4)
  1. [§3.2, Eq. (23)] The central premise of the framework is that switch detection recovers the correct number and locations of switches, because optimization is performed interval-by-interval on the detected segmentation. The noise-robustness study does not test this premise in the unknown-N_s case: the text states that the authors manually set the number of switches to N_s=1, so Figure 6 reports errors only for a known-switch-count, best-case scenario. The following sentence concedes that when N_s is unknown, switch detection algorithms often mistake noise jumps for switches, leading to unreliable parameters and untrustworthy switch locations, and Eq. (23) formalizes the resulting cascade. Since the abstract and introduction claim a solution to the general problem in Eqs. (2)-(3), this is a load-bearing gap: either the claims must be restricted to the known-N_s or user-tuned-σ setting, or the paper must add experiments in which N_s is estimated rather than supplied.
  2. [§3.3, Eqs. (24)-(27)] The non-uniform switching example reports the algorithmic configuration σ=N_s=1, but the true switch set for the system is {5,10,15,20}, i.e., four switches: α_m switches at 5 and 15, while α_p switches at 5, 10, 15, and 20, and the union contains four distinct locations. Under the description of Algorithm 1, setting σ=N_s=1 would force exactly one detected switch or trigger the error at line 24 when the detected count differs from the specified N_s. The reported H_s=0 and E_Ns=0 are therefore not reproducible from the stated configuration. Please correct the stated σ or clarify its meaning, and verify that the non-uniform switching example actually used an unknown-N_s detection mode; this is load-bearing because the example is used to support the non-uniform switching claim.
  3. [§3.4, p. 15] The sparse-regression advection-diffusion example contains conflicting error reports. Immediately after the algorithmic setup, the text says 'we achieved E_t=0.001 and E_p=0.001'; two paragraphs later it says that without sparse regression the parameter error is E_p=0.16 and with sparse regression E_p=0.05. These are different claims about the same experiment. If E_p=0.05 is the correct sparse-regression result, the earlier E_p=0.001 must be corrected; as written, the reader cannot determine which metric corresponds to which configuration, and the central demonstration of the sparse-regression extension is undermined.
  4. [§3.4, Eq. (30), Fig. 8C] The estimated parameter α̂(t)=0.97 sin(3t+11.5)+ε is not a valid phase reconstruction of α(t)=sin(3t-1). The phase difference is 12.5 radians (or approximately 6.22 radians after reducing modulo 2π), not approximately -1 radian, so the statement that the -1 radian offset is approximately equal to the 11.5 radian offset is mathematically incorrect. Because this example is the main evidence that continuous parameter functions can be recovered by sparse regression, the phase discrepancy must either be reported honestly as a limitation or the estimate must be recomputed with a corrected dictionary/phase handling.
minor comments (7)
  1. [§2.2, Algorithm 1, lines 11-14] The pseudocode first overwrites the switch list with the result and then checks, for each entry, whether that entry is already in the list; this condition is tautologically true and cannot perform the intended cross-state deduplication. Please rewrite lines 11-14 so the redundancy-removal step is well defined.
  2. [§2.4, Eq. (8)] Equation (8) uses an ℓ0 penalty, while the text and the advection-diffusion example refer to LASSO and implement λ∥w∥1. Please clarify which objective is actually minimized and whether ℓ1 or ℓ0 regularization was used in the reported experiments.
  3. [§2.4] The choice of n=N/6 segments for continuously varying parameters is stated without justification or sensitivity analysis; please explain the rationale and report how the results depend on this choice.
  4. [§3.1, Eq. (14)] The notation D(t)=p1(t) t p2(t) is ambiguous; it should presumably read p1(t) t^{p2(t)}, and the text would benefit from an explicit statement of the intended exponentiation.
  5. [§3.3, Fig. 7B] The time-interval labels in Figure 7B are not in chronological order; the intervals should be ordered [0,5], (5,10], (10,15], (15,20], (20,25] for readability.
  6. [Throughout] There are numerous typographical errors, including 'swtich', 'recrusively', 'challaenged', 'nxy and kxy', and 'utlize'; a careful proofread is needed.
  7. [§1.1, Definition 2] The definition introduces continuous switch locations and the continuous switch number N_sc, but no experiment estimates such locations; please state explicitly whether the continuous-switch part of the definition is intended to be operationalized in future work.

Circularity Check

1 steps flagged · score 6.0 of 10

Non-uniform-switching validation feeds N_s=1 into binseg while the model defines four switches, so the reported E_Ns=0 and H_s=0 are not independent estimates; the rest of the framework is benchmarked against synthetic ground truth and is largely non-circular.

  1. fitted input called prediction [Section 3.3, Equations (24)-(27), Algorithm 1, and Table 1 metric definitions]
    "αm(t)=4.0 if t∈[0,5]; 5.0 if t∈(5,15]; 6.0 if t∈(15,24]; αp(t)=4.0 if t∈[0,5]; 5.0 if t∈(5,10]; 6.0 if t∈(10,15]; 7.0 if t∈(15,20]; 8.0 if t∈(20,25] ... binseg for switch detection with auto-regressive cost, σ=N s = 1, sg = 10 ... With this algorithmic set-up, we achieved Hs = 0, ENs = 0, Et = 0.001 and Ep = 0.05."

    The model in Eqs. (24)-(27) has four switch locations (5, 10, 15, 20) across αm and αp, so true N_s=4. The stated setup instead passes σ=N_s=1 to binseg, which per Eq. (5) fixes the number of switches to one. The reported E_Ns=0 and H_s=0 therefore cannot be an inferred validation of switch detection: with the quoted input, any detected set has one switch, and E_Ns=N_s_true−N̂_s would be 3, not 0. The reported metrics are consistent only if the 'true' N_s used in the evaluation is taken to be the input value 1, making the switch-number 'prediction' self-definitional rather than a test against the stated ground truth. This example thus presents a user-supplied switch count as an estimated success.

full rationale

The framework itself is an optimization pipeline evaluated on synthetic data with known ground-truth parameters and switch times; the reported parameter and trajectory errors (E_p, E_t) compare estimates against true functions, so they are not circular by construction. Self-citations ([16], [56]) are used only as background or to reference the PVTS model whose equations are restated in the paper; they are not load-bearing. The sparse-regression step fits a dictionary to parameter samples obtained from the data and is then scored against the true α(t), which is an external benchmark. The one genuine circularity is in the non-uniform-switching validation: with σ=N_s=1 passed to binseg while the stated true model has four switches, the reported E_Ns=0 and H_s=0 are not independent estimates; they are consistent only if the metric's 'true' N_s is taken to be the input N_s=1. The noise-robustness section transparently states that the noisy-data study fixes N_s=1, so that is a limitation rather than a hidden circularity. Overall, the central claim has independent support from the PVTS and heat-equation examples, but the non-uniform-switching 'prediction' reduces by construction, giving a partial-circularity score of 6.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The framework rests on known model structure, user-tuned hyperparameters (sigma, sg, segment count, lambda, bounds, optimizer choice), and the reliability of the switch detection stage. No new physical entities are introduced.

free parameters (7)
  • binseg noise level or switch count sigma = 10^-5, 10^-3, Ns=1, N/6 across examples
    User-chosen hyperparameter controlling binseg sensitivity; the paper calls sigma the most important hyperparameter for switch detection (Section 2.2).
  • switch gap sg = 5, 10, or 1 across examples
    Minimum number of data points between detected switches, set per example (Section 2.2).
  • binseg cost function = auto-regressive cost in all examples
    Choice of change-point cost function; the paper lists many options but does not justify the auto-regressive selection.
  • sparse regression segment count N/6 = N/6, e.g. 16 segments for N=100
    Arbitrary segmentation of data to sample the continuous parameter before LASSO; stated without justification in Section 2.4.
  • LASSO regularization lambda = 0.01
    Set for the advection-diffusion example; no tuning procedure is specified (Section 3.4).
  • optimizer bounds for advection-diffusion = |alpha|<10, |D|<10
    User-prescribed bounds used with Nelder-Mead in Section 3.4.
  • optimizer selection = Differential Evolution for PVTS, Nelder-Mead elsewhere
    Different optimizers are used across examples, and the results depend on this choice (Sections 3 and 3.1).
assumptions (5)
  • domain assumption The model structure f(X, p(t)) is fully known a priori; only the parameters are unknown.
    Problem formulation in Section 1.2 assumes the model form is given, which is central to the segmented optimization approach.
  • domain assumption Discrete parameter switches appear as non-differentiable points in the trajectory X(t), and binseg can reliably detect them from sampled data.
    Remark 1 and Section 2.2; the entire switch detection stage depends on this, and Section 3.2 shows noise can break it.
  • domain assumption Numerical integration during optimization yields sufficiently accurate X_model.
    Remark 2 assumes integration error below a desired epsilon; the paper notes LSODA performed worse than Forward Euler in one example, showing solver choice matters.
  • ad hoc to paper The sparse regression dictionary spans the true continuous parameter function.
    In Section 3.4, the recovered expression includes sin(3t+11.5), e^{-2.55 t^0.6}, and t^0.96, implying a hand-chosen dictionary whose full contents are not specified.
  • domain assumption Global optimization algorithms such as Differential Evolution locate global minima of the non-convex objective over the chosen bounds.
    Remark 3 states this conditionally on using a global optimizer; the PVTS example relies on this without a certificate of global optimality.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Estimating Varying Parameters in Dynamical Systems: A Modular Framework Using Switch Detection, Optimization, and Sparse Regression." pith.science (2026). https://pith.science/paper/JQKO2OYP

@misc{pith2026241216198,
  author       = {Pith},
  title        = {Pith review of: Estimating Varying Parameters in Dynamical Systems: A Modular Framework Using Switch Detection, Optimization, and Sparse Regression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JQKO2OYP}},
  note         = {Machine review of arXiv:2412.16198}
}
read the original abstract

The estimation of static parameters in dynamical systems and control theory has been extensively studied, with significant progress made in estimating varying parameters in specific system types. Suppose, in the general case, we have data from a system with parameters that depend on an independent variable such as time or space. Further, suppose the system's model structure is known, but our aim is to identify functions describing parameter-varying elements as they change with respect to time or another variable. Focusing initially on the subclass of problems where parameters are discretely switching piecewise constant functions, we develop an algorithmic framework for detecting discrete parameter switches and fitting a piecewise constant model to data using optimization-based parameter estimation. Our modular framework allows for customization of switch detection, numerical integration, and optimization sub-steps to suit user requirements. Binary segmentation is used for switch detection, with Nelder-Mead and Powell methods employed for optimization. To address broader problems, we extend our framework using dictionary-based sparse regression with trigonometric and polynomial functions to obtain continuously varying parameter functions. Finally, we assess the framework's robustness to measurement noise. We demonstrate its capabilities across several examples, including time-varying promoter-gene expression, a genetic toggle switch, a parameter-switching manifold, the heat equation with a time-varying diffusion coefficient, and the advection-diffusion equation with a continuously varying parameter.

Figures

Figures reproduced from arXiv: 2412.16198 by the authors.

Figure 1
Figure 1. Data Model Segmented Data Parameters (piece-wise constant) Switch Detection Optimization Numerical Integration Sparse Regression Parameters (continuous) Model Trajectory Xdata dXmodel dt = f(X , p(t) , pstatic) min p(t) Xdatai − Xmodeli 2 Xmodeli = ∫ t i ti−1 dXmodel dt dt Xdata1 Xdatan [ | | ] [t 0, t 1] … (t n−1, t n] [p1 |⋯|pn] θw min θ,w p − θw 2 2 + λ∥w∥0 Xmodel1 Xmodeln [ | | ] Legend A priori input Output Mod… view at source ↗
Figure 2
Figure 2. Outline of algorithm with sparse regression for the identification of continuously varying parameters. [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. (A) Solution to parameter varying toggle switch and reconstruction of solution with estimated parameters. (B) [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: (A) True solution to heat equation PDE. (B) Reconstruction of solution to heat equation PDE with estimated [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: (A) Synthetic data and reconstruction from estimated parameters. (B) Synthetic data with white noise [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: (A) Hausdorff error vs. standard deviation of noise. (B) Parameter error vs. standard deviation of noise. (C) [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: (A) Trajectory of hybrid promoter model with different parameters ( [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: (A) Synthetic data from advection-diffusion equation containing a continuously varying parameter. The [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Variational Parameter Calibration with Physics-Aware Latent-Space Surrogates

    cs.LG 2026-08 conditional novelty 6.0 of 10

    Observable-supervised autoencoder surrogates improve variational parameter calibration for CFD flows compared with reconstruction-only surrogate baselines.

Reference graph

Works this paper leans on

67 extracted references · 61 canonical work pages · cited by 1 Pith paper

  1. [1]

    Ryan and A

    E.P. Ryan and A. Ichikawa. Parameter estimation and control for distributed systems.IFAC Proceedings Volumes, 12(7):149–155, 1979. IFAC Symposium on computer Aided Design of Control Systems, Zurich, Switzerland, 29-31 August

    work page 1979

  2. [2]

    Burns, Eugene M

    John A. Burns, Eugene M. Cliff, and Kasie Farlow. Parameter estimation and model discrepancy in control systems with delays.IFAC Proceedings Volumes, 47(3):11679–11684, 2014. 19th IFAC World Congress

    work page 2014

  3. [3]

    Parameter estimation and uncertainty quantification for systems biology models.Current opinion in systems biology, Dec 2019

    Eshan D Mitra and William S Hlavacek. Parameter estimation and uncertainty quantification for systems biology models.Current opinion in systems biology, Dec 2019

    work page 2019

  4. [4]

    Penas, Patricia González, Jose A

    David R. Penas, Patricia González, Jose A. Egea, Ramón Doallo, and Julio R. Banga. Parameter estimation in large-scale systems biology models: A parallel and self-adaptive cooperative strategy.BMC Bioinformatics, 18(1), Jan 2017

    work page 2017

  5. [5]

    Parameter estimation in systems biology models using spline approximation

    Choujun Zhan and Lam F Yeung. Parameter estimation in systems biology models using spline approximation. BMC Systems Biology, 5(1):14, 2011

    work page 2011

  6. [6]

    Valderrama-Bahamóndez and Holger Fröhlich

    Gloria I. Valderrama-Bahamóndez and Holger Fröhlich. Mcmc techniques for parameter estimation of ode based models in systems biology.Frontiers in Applied Mathematics and Statistics, 5, Nov 2019

    work page 2019

  7. [7]

    Robust parameter estimation and identifiability analysis with hybrid neural ordinary differential equations in computational biology

    Stefano Giampiccolo, Federico Reali, Anna Fochesato, Giovanni Iacca, and Luca Marchetti. Robust parameter estimation and identifiability analysis with hybrid neural ordinary differential equations in computational biology. Robust parameter estimation and identifiability analysis with hybrid neural ordinary differential equations in computational biology, Jun 2024

    work page 2024

  8. [8]

    Kaandorp, and Joke G

    Maksat Ashyraliyev, Yves Fomekong-Nanfack, Jaap A. Kaandorp, and Joke G. Blom. Systems biology: parameter estimation for biochemical models.The FEBS Journal, 276(4):886–902, 2009

    work page 2009

Show all 67 references
  1. [9]

    Douspis, J

    M. Douspis, J. G. Bartlett, A. Blanchard, and M. Le Dour. Concerning parameter estimation using the cosmic microwave background.Astronomy and Astrophysics, 368(1):1–14, Mar 2001

    2001

  2. [10]

    Bailer-Jones.Astronomical Object Classification and Parameter Estimation with the Gaia Galactic Survey Satellite

    Coryn A.L. Bailer-Jones.Astronomical Object Classification and Parameter Estimation with the Gaia Galactic Survey Satellite. Springer Berlin Heidelberg, Berlin, Heidelberg, 2005

    2005

  3. [11]

    Switch detection and robust parameter estimation for slowly switched hammerstein systems.Nonlinear Analysis: Hybrid Systems, 32:202–213, 2019

    Zhu Wang, Haoran An, and Xionglin Luo. Switch detection and robust parameter estimation for slowly switched hammerstein systems.Nonlinear Analysis: Hybrid Systems, 32:202–213, 2019

    2019

  4. [12]

    Identification of switched dynamic system for electric multiple unit train modeling.Control Engineering Practice, 143:105815, 2024

    Wenju Zheng, Yiming Wan, Fan Yang, Chao Shang, Hao Ye, Ming Jiang, and Jia Wang. Identification of switched dynamic system for electric multiple unit train modeling.Control Engineering Practice, 143:105815, 2024

    2024

  5. [13]

    A randomized method for the identification of switched narx systems.Nonlinear Analysis: Hybrid Systems, 49:101364, 2023

    Miao Yu, Federico Bianchi, and Luigi Piroddi. A randomized method for the identification of switched narx systems.Nonlinear Analysis: Hybrid Systems, 49:101364, 2023

    2023

  6. [14]

    Model structure selection for switched narx system identification: A randomized approach.Automatica, 125:109415, 2021

    Federico Bianchi, Valentina Breschi, Dario Piga, and Luigi Piroddi. Model structure selection for switched narx system identification: A randomized approach.Automatica, 125:109415, 2021

    2021

  7. [15]

    A novel system identification algorithm for nonlinear markov jump system

    Haoyu Li, Ke Zhang, and Minghu Tan. A novel system identification algorithm for nonlinear markov jump system. Information Sciences, 616:348–366, 2022

    2022

  8. [16]

    Nonlinear data-driven control via koopman models: the interplay of stabilization and subspace closure error.arXiv e-prints, 2024

    Charles A Johnson and Enoch Yeung. Nonlinear data-driven control via koopman models: the interplay of stabilization and subspace closure error.arXiv e-prints, 2024. 17 arXivA PREPRINT

    2024

  9. [17]

    Control of oscillatory temperature field in a building via damping assignment to nonlinear koopman mode

    Yoshihiko Susuki, Kohei Eto, Naoto Hiramatsu, and Atsushi Ishigame. Control of oscillatory temperature field in a building via damping assignment to nonlinear koopman mode. In2022 IEEE Conference on Control Technology and Applications (CCTA), pages 796–801. IEEE, 2022

    2022

  10. [18]

    Control of in-room temperature field via damping assignment to nonlinear koopman mode.IEEE Transactions on Control Systems Technology, 2024

    Yoshihiko Susuki, Kohei Eto, Naoto Hiramatsu, and Atsushi Ishigame. Control of in-room temperature field via damping assignment to nonlinear koopman mode.IEEE Transactions on Control Systems Technology, 2024

    2024

  11. [19]

    Direct learning for parameter-varying feedforward control: A neural-network approach, 2023

    Johan Kon, Jeroen van de Wijdeven, Dennis Bruijnen, Roland Tóth, Marcel Heertjes, and Tom Oomen. Direct learning for parameter-varying feedforward control: A neural-network approach, 2023

    2023

  12. [20]

    Parameter estimation for a class of time-varying systems with the invariant matrix

    Ning Xu and Feng Ding. Parameter estimation for a class of time-varying systems with the invariant matrix. International Journal of Robust and Nonlinear Control, 33(3):2163–2181, 2023

    2023

  13. [21]

    Parameter estimation on linear time-varying systems.Journal of the Franklin Institute, 348(4):777–789, 2011

    Luiz Cláudio Andrade Souza and Reinaldo Martinez Palhares. Parameter estimation on linear time-varying systems.Journal of the Franklin Institute, 348(4):777–789, 2011

    2011

  14. [22]

    Johnson, Stefano Di Cairano, and Ricardo G

    Ryan S. Johnson, Stefano Di Cairano, and Ricardo G. Sanfelice. Parameter estimation for hybrid dynamical systems using hybrid gradient descent. In2021 60th IEEE Conference on Decision and Control (CDC), pages 4648–4653, 2021

    2021

  15. [23]

    Bayesian parameter estimation for linear parameter-varying systems with outliers and missing observations.IEEE Access, pages 1–1, 2024

    Liming Qian. Bayesian parameter estimation for linear parameter-varying systems with outliers and missing observations.IEEE Access, pages 1–1, 2024

    2024

  16. [24]

    Fatimah Al-Taie, Ibtisam Kamil Hanan, and Fadhel S. Fadhel. Extended robust controller design for parameter varying multi-dimensional systems.AIP Conference Proceedings, 3079(1):050005, 04 2024

    2024

  17. [25]

    Selective review of offline change point detection methods

    Charles Truong, Laurent Oudre, and Nicolas Vayatis. Selective review of offline change point detection methods. Signal Processing, 167:107299, 2020

    2020

  18. [26]

    A more efficient algorithm to compute the rand index for change-point problems, 2021

    Lucas de Oliveira Prates. A more efficient algorithm to compute the rand index for change-point problems, 2021

    2021

  19. [27]

    Sarmin and L.A

    E.N. Sarmin and L.A. Chudov. On the stability of the numerical integration of systems of ordinary differential equations arising in the use of the straight line method.USSR Computational Mathematics and Mathematical Physics, 3(6):1537–1543, 1963

    1963

  20. [28]

    Ryan Prescott Adams and David J. C. MacKay. Bayesian online changepoint detection, 2007

    2007

  21. [29]

    Fearnhead R

    P. Fearnhead R. Killick and I. A. Eckley. Optimal detection of changepoints with a linear computational cost. Journal of the American Statistical Association, 107(500):1590–1598, 2012

    2012

  22. [30]

    Rudin, Stanley Osher, and Emad Fatemi

    Leonid I. Rudin, Stanley Osher, and Emad Fatemi. Nonlinear total variation based noise removal algorithms. Physica D: Nonlinear Phenomena, 60(1):259–268, 1992

    1992

  23. [31]

    E. S. PAGE. CONTINUOUS INSPECTION SCHEMES.Biometrika, 41(1-2):100–115, 06 1954

    1954

  24. [32]

    Jackson, J.D

    B. Jackson, J.D. Scargle, D. Barnes, S. Arabhi, A. Alt, P. Gioumousis, E. Gwin, P. Sangtrakulcharoen, L. Tan, and Tun Tao Tsai. An algorithm for optimal partitioning of data on an interval.IEEE Signal Processing Letters, 12(2):105–108, 2005

    2005

  25. [33]

    L.R. Rabiner. A tutorial on hidden markov models and selected applications in speech recognition.Proceedings of the IEEE, 77(2):257–286, 1989

    1989

  26. [34]

    Retrospective mutiple change-point estimation with kernels.2007 IEEE/SP 14th Workshop on Statistical Signal Processing, pages 768–772, 2007

    Zaïd Harchaoui and Olivier Cappé. Retrospective mutiple change-point estimation with kernels.2007 IEEE/SP 14th Workshop on Statistical Signal Processing, pages 768–772, 2007

    2007

  27. [35]

    Mallat.A wavelet tour of signal processing

    S. Mallat.A wavelet tour of signal processing. Science Direct, 1999

    1999

  28. [36]

    Fast detection of multiple change-points shared by many signals using group lars

    Jean-philippe Vert and Kevin Bleakley. Fast detection of multiple change-points shared by many signals using group lars. In J. Lafferty, C. Williams, J. Shawe-Taylor, R. Zemel, and A. Culotta, editors,Advances in Neural Information Processing Systems, volume 23. Curran Associa...

    2010

  29. [37]

    Change-point detection in time-series data by relative density-ratio estimation.Neural Networks, 43:72–83, 2013

    Song Liu, Makoto Yamada, Nigel Collier, and Masashi Sugiyama. Change-point detection in time-series data by relative density-ratio estimation.Neural Networks, 43:72–83, 2013

    2013

  30. [38]

    Lavielle and G

    M. Lavielle and G. Teyssière. Detection of multiple change-points in multivariate time series.Lithuanian Mathematical Journal, 46(3):287–306, Jul 2006

    2006

  31. [39]

    M Lee Richard A Davis and Gabriel A Rodriguez-Yam

    Thomas C. M Lee Richard A Davis and Gabriel A Rodriguez-Yam. Structural break estimation for nonstationary time series models.Journal of the American Statistical Association, 101(473):223–239, 2006

    2006

  32. [40]

    Estimating multiple breaks one at a time.Econometric Theory, 13(3):315–352, 1997

    Jushan Bai. Estimating multiple breaks one at a time.Econometric Theory, 13(3):315–352, 1997

    1997

  33. [41]

    Wild binary segmentation for multiple change-point detection.The Annals of Statistics, 42(6):2243 – 2281, 2014

    Piotr Fryzlewicz. Wild binary segmentation for multiple change-point detection.The Annals of Statistics, 42(6):2243 – 2281, 2014. 18 arXivA PREPRINT

    2014

  34. [42]

    Savitzky and M

    Abraham. Savitzky and M. J. Golay. Smoothing and differentiation of data by simplified least squares procedures. Analytical Chemistry, 36(8):1627–1639, Jul 1964

    1964

  35. [43]

    Why and how savitzky–golay filters should be replaced.ACS Measurement Science Au, 2(2):185–196, 2022

    Michael Schmid, David Rath, and Ulrike Diebold. Why and how savitzky–golay filters should be replaced.ACS Measurement Science Au, 2(2):185–196, 2022

    2022

  36. [44]

    E. T. Whittaker. On a new method of graduation.Proceedings of the Edinburgh Mathematical Society, 41:63–75, 1922

    1922

  37. [45]

    J. A. Nelder and R. Mead. A Simplex Method for Function Minimization.The Computer Journal, 7(4):308–313, 01 1965

    1965

  38. [46]

    M. J. D. Powell. An efficient method for finding the minimum of a function of several variables without calculating derivatives.The Computer Journal, 7(2):155–162, 01 1964

    1964

  39. [47]

    Attila Gábor and Julio R. Banga. Robust and efficient parameter estimation in dynamic models of biological systems.BMC Systems Biology, 9(1), Oct 2015

    2015

  40. [48]

    Wales and Jonathan P

    David J. Wales and Jonathan P. Doye. Global optimization by basin-hopping and the lowest energy structures of lennard-jones clusters containing up to 110 atoms.The Journal of Physical Chemistry A, 101(28):5111–5116, Jul 1997

    1997

  41. [49]

    Springer Berlin Heidelberg, Berlin, Heidelberg, 2004

    Jouni Lampinen and Rainer Storn.Differential Evolution, pages 123–166. Springer Berlin Heidelberg, Berlin, Heidelberg, 2004

    2004

  42. [50]

    A double annealing algorithm for discrete location/allocation problems.European Journal of Operational Research, 86(3):452–468, 1995

    Giovanni Righini. A double annealing algorithm for discrete location/allocation problems.European Journal of Operational Research, 86(3):452–468, 1995

    1995

  43. [51]

    Endres, Carl Sandrock, and Walter W

    Stefan C. Endres, Carl Sandrock, and Walter W. Focke. A simplicial homology algorithm for lipschitz optimisation. Journal of Global Optimization, 72(2):181–217, Mar 2018

    2018

  44. [52]

    D. R. Jones, C. D. Perttunen, and B. E. Stuckman. Lipschitzian optimization without the lipschitz constant. Journal of Optimization Theory and Applications, 79(1):157–181, Oct 1993

    1993

  45. [53]

    Sparse signal reconstruction on fixed and adaptive supervised dictionary learning for transient stability assessment.Energies, 14(23), 2021

    Raoult Teukam Dabou, Innocent Kamwa, Jacques Tagoudjeu, and Francis Chuma Mugombozi. Sparse signal reconstruction on fixed and adaptive supervised dictionary learning for transient stability assessment.Energies, 14(23), 2021

    2021

  46. [54]

    A Tudosa

    L De Vito, E Picariello, F Picariello, S Rapuano, and I. A Tudosa. Dictionary optimization method for reconstruc- tion of ecg signals after compressed sensing.Sensors (Basel), 2021

    2021

  47. [55]

    Signal processing using dictionaries, atoms, and deep learning: A common analysis-synthesis framework.Proceedings of the IEEE, 110(4):454–475, 2022

    Chao Zhang and Mirko Van Der Baan. Signal processing using dictionaries, atoms, and deep learning: A common analysis-synthesis framework.Proceedings of the IEEE, 110(4):454–475, 2022

    2022

  48. [56]

    Stability analysis of parameter varying genetic toggle switches using koopman operators.Mathematics, 9(23), 2021

    Jamiree Harrison and Enoch Yeung. Stability analysis of parameter varying genetic toggle switches using koopman operators.Mathematics, 9(23), 2021

    2021

  49. [57]

    Learning noise-robust stable koopman operator for control with physics- informed observables.arXiv e-prints, pages arXiv–2408, 2024

    Shahriar Akbar Sakib and Shaowu Pan. Learning noise-robust stable koopman operator for control with physics- informed observables.arXiv e-prints, pages arXiv–2408, 2024

    2024

  50. [58]

    Learning noise-robust stable koopman operator for control with hankel dmd.arXiv preprint arXiv:2408.06607, 2024

    Shahriar Akbar Sakib and Shaowu Pan. Learning noise-robust stable koopman operator for control with hankel dmd.arXiv preprint arXiv:2408.06607, 2024

    2024 arXiv

  51. [59]

    Online real-time learning of dynamical systems from noisy streaming data.Scientific Reports, 13(1):22564, 2023

    Subhrajit Sinha, Sai P Nandanoori, and DA Barajas-Solano. Online real-time learning of dynamical systems from noisy streaming data.Scientific Reports, 13(1):22564, 2023

    2023

  52. [60]

    Boddupalli, T

    N. Boddupalli, T. Matchen, J. Moehlis Symbolic regression via neural networksChaos, 33(8), 2023

    2023

  53. [61]

    Adarsh Iyer, Nibodh Boddupalli, Jeff Moehlis Expressive Symbolic Regression for Interpretable Models of Discrete-Time Dynamical SystemsarXiv preprint arXiv:2406.06585, 2024

    2024 arXiv

  54. [62]

    Martinelli

    A. Martinelli. Identifiability of nonlinear ODE models with time-varying parameters: the general analytical solution and applications in viral dynamics.arXiv preprint arXiv:2211.13507, 2024. https://arxiv.org/abs/ 2211.13507

    2024 arXiv

  55. [63]

    A. F. Villaverde, A. Barreiro, and A. Papachristodoulou. Structural Identifiability of Dynamic Systems Biology Models.PLOS Computational Biology, 12(10):e1005153, 2016. https://doi.org/10.1371/journal.pcbi. 1005153

    2016 doi

  56. [64]

    Villaverde, N

    A.F. Villaverde, N. Tsiantis, and J.R. Banga. Full observability and estimation of unknown inputs, states and parameters of nonlinear biological models.Journal of the Royal Society Interface, 16(156):20190043, 2019. doi:10.1098/rsif.2019.0043. 19 arXivA PREPRINT

    2019

  57. [65]

    S.Z. Yong, M. Zhu, and E. Frazzoli. Simultaneous input and state estimation for linear time-varying continuous-time stochastic systems.IEEE Transactions on Automatic Control, 62(5):2531–2538, 2017. doi:10.1109/TAC.2016.2601688

    2017

  58. [66]

    Curi, K.Y

    S. Curi, K.Y . Levy, and A. Krause. Adaptive input estimation in linear dynamical systems with applications to learning-from-observations. InProceedings of the 58th IEEE Conference on Decision and Control (CDC), pages 4115–4120, Nice, France, 2019. doi:10.1109/CDC40024.2019.9029379

    2019

  59. [67]

    S. L. Brunton, J. L. Proctor, and J. N. Kutz. Discovering governing equations from data by sparse identification of nonlinear dynamical systems.Proceedings of the National Academy of Sciences, 113(15):3932–3937, 2016. doi:10.1073/pnas.1517384113. 20

    2016 doi

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.