pith. sign in

arxiv: 2605.18292 · v1 · pith:EWUB5J5Inew · submitted 2026-05-18 · 📡 eess.SY · cs.SY

Learning the dynamics of nonlinear systems with regional stability guarantees through linear matrix inequality constraints

Daniel Frank , Fahim Shakib , Steffen Staab This is my paper

Pith reviewed 2026-05-20 09:35 UTC · model grok-4.3

classification 📡 eess.SY cs.SY
keywords regional stabilityrecurrent neural networkslinear matrix inequalitiesbarrier functionsdeadzone activationforward invariancesystem identificationnonlinear dynamics
0
0 comments X

The pith

Recurrent neural networks can be learned from regional data with a certificate of stability inside a chosen compact state set.

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

The paper develops a training procedure for recurrent neural networks that produces models with a built-in guarantee of staying inside a user-specified compact region of the state space whenever the input stays inside a given set. It achieves this by first establishing generalized sector conditions on a deadzone nonlinearity that make the region forward invariant, then converting those conditions into linear matrix inequalities through a barrier function. The resulting constraints are added to the learning objective so that the optimized network inherits a stability certificate only where data was actually collected. This regional focus matters because it avoids the over-restriction of global stability demands that frequently prevent accurate fits to limited observations.

Core claim

By relying on generalized sector conditions on the deadzone activation function, sufficient conditions are derived that guarantee forward invariance on a compact set of the state space for any inputs from a given set. A barrier function approach then yields linear matrix inequality constraints that can be imposed during learning, producing recurrent neural network models equipped with a certificate of regional stability in a subset of the state space and for a given input set.

What carries the argument

Generalized sector conditions on the deadzone activation function combined with a barrier function that together produce linear matrix inequality constraints enforcing forward invariance and regional stability.

If this is right

  • The learned models match observed input-output behavior while remaining provably stable inside the data region.
  • Global stability constraints often fail to identify the underlying system from regional observations due to excessive conservatism.
  • Unconstrained learning produces models without any stability guarantee even inside the observed region.
  • The certificate applies only to the chosen compact state subset and input set, aligning with the limited scope of the data.

Where Pith is reading between the lines

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

  • The same regional certificate could be reused to design controllers that keep trajectories inside the certified set.
  • Similar barrier constructions might extend the approach to other activation functions or deeper network structures.
  • The linear matrix inequality form suggests the method can be combined with convex solvers for larger state dimensions.

Load-bearing premise

The unknown dynamical system that generated the data satisfies the forward invariance properties implied by the generalized sector conditions on the deadzone activation function for the chosen compact set and input set.

What would settle it

A concrete counter-example would be a learned model that satisfies the linear matrix inequalities yet produces a trajectory that exits the compact state set or loses stability for some input inside the given input set.

Figures

Figures reproduced from arXiv: 2605.18292 by Daniel Frank, Fahim Shakib, Steffen Staab.

Figure 1
Figure 1. Figure 1: Model (2) with static nonlinear function ∆. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Nonlinear activation functions. 3. REGIONAL STABILITY ANALYSIS FOR NONLINEAR SYSTEM IDENTIFICATION The nonlinear model (2) is an interconnection between a linear model and a static, memoryless, and sector-bounded nonlinear function. In the literature, these models are known as Lur’e-type systems; see, for example, (Khalil, 2002; Shakib et al., 2022; Scherer, 2022). To assess their stability properties, sec… view at source ↗
Figure 3
Figure 3. Figure 3: Predictions made by the learned models in phase space, with excitation from the test dataset. The red crosses [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Predictions of the models in the time domain compared to the outputs from the test dataset. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
read the original abstract

This paper presents a method that learns a regionally stable recurrent neural network model from a set of input-output data generated by an unknown dynamical system. Relying on generalized sector conditions on the deadzone activation function, we first derive sufficient conditions that guarantee forward invariance on a compact set of the state space for any inputs from a given set. Such regional properties lead to less conservative conditions compared to variants that offer a global form of stability, and are in line with the system data that is only observed regionally. Our learning method derives conditions for regional stability using a barrier function approach, leading to models equipped with a certificate of regional stability in a subset of the state space and for a given input set. We illustrate our theoretical result with a numerical example and compare it to methods that impose a global form of stability, which fail to identify the system, and with a method that imposes no stability constraints at all, which does not guarantee a stable behavior within any state or input set.

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.

Referee Report

1 major / 2 minor

Summary. The paper proposes a method to learn recurrent neural network models from input-output data of an unknown nonlinear dynamical system, deriving sufficient LMI conditions based on generalized sector bounds for a deadzone activation function and a barrier function approach to guarantee forward invariance (regional stability) of the learned model on a user-chosen compact state set for inputs from a specified set. It claims these regional conditions are less conservative than global stability variants, and illustrates the approach with a numerical example comparing against global-stability-constrained and unconstrained learning.

Significance. If the LMI derivations are sound and the regional sets can be reliably selected, the work could provide a useful framework for learning provably regionally stable models from limited regional data, with the barrier-function and sector-based LMI approach offering a concrete certificate mechanism that aligns with observed trajectories. The numerical comparison highlighting failure modes of global constraints is a practical strength.

major comments (1)
  1. Theoretical result section: The sufficient LMI conditions for forward invariance of the learned RNN are derived under the assumption that the unknown data-generating system itself satisfies the forward invariance properties implied by the generalized sector conditions on the deadzone nonlinearity for the chosen compact state set and input set. No mechanism is provided to certify or select such sets from data alone; if this assumption does not hold for the true system, the regional certificate on the learned model becomes disconnected from the observed data, which is load-bearing for the central claim of obtaining models 'equipped with a certificate of regional stability' from data.
minor comments (2)
  1. Abstract and introduction: Clarify whether the barrier function is used only for the certificate or also in the learning objective, as the current wording leaves the optimization formulation ambiguous.
  2. Numerical example section: Provide more detail on how the compact state and input sets were chosen for the example, and report the specific LMI feasibility margins or solver tolerances to allow reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback and positive assessment of the work's potential utility. We address the single major comment below with clarifications on the scope of the theoretical results and indicate where revisions will be made.

read point-by-point responses

  1. Referee: Theoretical result section: The sufficient LMI conditions for forward invariance of the learned RNN are derived under the assumption that the unknown data-generating system itself satisfies the forward invariance properties implied by the generalized sector conditions on the deadzone nonlinearity for the chosen compact state set and input set. No mechanism is provided to certify or select such sets from data alone; if this assumption does not hold for the true system, the regional certificate on the learned model becomes disconnected from the observed data, which is load-bearing for the central claim of obtaining models 'equipped with a certificate of regional stability' from data.

    Authors: We thank the referee for this observation. We clarify that the generalized sector conditions are applied to the deadzone activation function of the learned RNN model, not to the unknown data-generating system. The LMI conditions are derived solely for the model to guarantee forward invariance of its trajectories on the user-chosen compact set for inputs from the specified set. These conditions hold by construction for the model once its parameters satisfy the LMIs, independent of the true system's behavior. The input-output data is used only to fit the model parameters while enforcing the LMIs; no assumption is made that the true system obeys the same sector bounds or invariance. The sets are selected by the user based on the region of data collection, consistent with the paper's statement that regional properties are 'in line with the system data that is only observed regionally.' We do not claim to certify the true system or provide an automated data-driven procedure for set selection or verification, as the contribution focuses on sufficient LMI conditions for the model once sets are specified. The certificate applies to the learned model, which remains useful even if the true system exhibits different behavior outside the observed region. We will add a clarifying paragraph in the theoretical results section to explicitly distinguish the model's certified properties from those of the data-generating system and discuss user selection of sets informed by data availability. revision: yes

Circularity Check

0 steps flagged

No circularity: LMI conditions derived independently from sector bounds and barrier functions

full rationale

The derivation begins with generalized sector conditions on the deadzone activation function to obtain sufficient LMI constraints guaranteeing forward invariance on a user-chosen compact set for inputs in a given set. A barrier function approach is then applied to produce regional stability certificates for the RNN model. These steps rely on standard Lyapunov-style analysis and sector inequalities that are mathematically independent of the specific data or learned parameters. The data-fitting step simply optimizes the RNN weights subject to the pre-derived LMIs; the certificate itself is not obtained by fitting or by renaming any input quantity. No self-citations, self-definitional loops, or fitted-input-as-prediction patterns appear in the chain. The explicit assumption that the unknown system obeys the same sector-based invariance is external to the derivation and does not create circularity within the paper's own equations.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard sector-bound assumptions for deadzone nonlinearities and the existence of a suitable barrier function; no new free parameters or invented entities are described in the abstract.

axioms (1)
  • domain assumption Generalized sector conditions hold for the deadzone activation function on the chosen compact set.
    Invoked to derive the LMI constraints that guarantee forward invariance.

pith-pipeline@v0.9.0 · 5701 in / 1276 out tokens · 35111 ms · 2026-05-20T09:35:11.408723+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

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

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

101 extracted references · 101 canonical work pages

  1. [1]

    IEEE Transactions on automatic control , volume=

    Multiobjective output-feedback control via LMI optimization , author=. IEEE Transactions on automatic control , volume=. 1997 , publisher=

    work page 1997

  2. [2]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Recurrent neural network controllers synthesis with stability guarantees for partially observed systems , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    work page

  3. [3]

    at-Automatisierungstechnik , volume=

    Robustness analysis and training of recurrent neural networks using dissipativity theory , author=. at-Automatisierungstechnik , volume=. 2022 , publisher=

    work page 2022

  4. [4]

    arXiv preprint arXiv:2212.05781 , year=

    Robust Recurrent Neural Network to Identify Ship Motion in Open Water with Performance Guarantees--Technical Report , author=. arXiv preprint arXiv:2212.05781 , year=

    work page arXiv

  5. [5]

    2022 IEEE 61st Conference on Decision and Control (CDC) , pages=

    Synthesis of Stabilizing Recurrent Equilibrium Network Controllers , author=. 2022 IEEE 61st Conference on Decision and Control (CDC) , pages=. 2022 , organization=

    work page 2022

  6. [6]

    IEEE Control Systems Letters , volume=

    A convex parameterization of robust recurrent neural networks , author=. IEEE Control Systems Letters , volume=. 2020 , publisher=

    work page 2020

  7. [7]

    2021 60th IEEE Conference on Decision and Control (CDC) , pages=

    Recurrent equilibrium networks: Unconstrained learning of stable and robust dynamical models , author=. 2021 60th IEEE Conference on Decision and Control (CDC) , pages=. 2021 , organization=

    work page 2021

  8. [8]

    IEEE Transactions on Automatic Control , volume=

    Recurrent equilibrium networks: Flexible dynamic models with guaranteed stability and robustness , author=. IEEE Transactions on Automatic Control , volume=. 2023 , publisher=

    work page 2023

  9. [9]

    Efficient and accurate estimation of

    Fazlyab, Mahyar and Robey, Alexander and Hassani, Hamed and Morari, Manfred and Pappas, George , journal=. Efficient and accurate estimation of

    work page

  10. [10]

    IEEE transactions on systems, man, and cybernetics , number=

    Neuronlike adaptive elements that can solve difficult learning control problems , author=. IEEE transactions on systems, man, and cybernetics , number=. 1983 , publisher=

    work page 1983

  11. [11]

    2004 , publisher=

    Convex optimization , author=. 2004 , publisher=

    work page 2004

  12. [12]

    Lecture Notes, Dutch Institute for Systems and Control, Delft, The Netherlands , volume=

    Linear matrix inequalities in control , author=. Lecture Notes, Dutch Institute for Systems and Control, Delft, The Netherlands , volume=

    work page

  13. [13]

    1995 , publisher=

    Artificial neural networks for modelling and control of non-linear systems , author=. 1995 , publisher=

    work page 1995

  14. [14]

    IFAC-PapersOnLine , volume=

    Absolute stability analysis of discrete time feedback interconnections , author=. IFAC-PapersOnLine , volume=. 2017 , publisher=

    work page 2017

  15. [15]

    Dissipativity, Convexity and Tight

    Scherer, Carsten W , journal=. Dissipativity, Convexity and Tight. 2022 , publisher=

    work page 2022

  16. [16]

    arXiv preprint arXiv:2306.00571 , year=

    Robust Exponential Stability and Invariance Guarantees with General Dynamic O'Shea-Zames-Falb Multipliers , author=. arXiv preprint arXiv:2306.00571 , year=

    work page arXiv

  17. [17]

    2013 , publisher=

    Nonlinear systems: analysis, stability, and control , author=. 2013 , publisher=

    work page 2013

  18. [18]

    IEEE Transactions on automatic control , volume=

    Losslessness, feedback equivalence, and the global stabilization of discrete-time nonlinear systems , author=. IEEE Transactions on automatic control , volume=. 1994 , publisher=

    work page 1994

  19. [19]

    IEEE transactions on neural networks and learning systems , volume=

    Multistep prediction of dynamic systems with recurrent neural networks , author=. IEEE transactions on neural networks and learning systems , volume=. 2019 , publisher=

    work page 2019

  20. [20]

    Kingma and Jimmy Ba , editor =

    Diederik P. Kingma and Jimmy Ba , editor =. Adam:. 3rd International Conference on Learning Representations,. 2015 , timestamp =

    work page 2015

  21. [21]

    2000 , publisher=

    Control-Oriented System Identification: An H Approach , author=. 2000 , publisher=

    work page 2000

  22. [22]

    1999 , edition=

    System Identification: Theory for the User , author=. 1999 , edition=

    work page 1999

  23. [23]

    2012 , publisher=

    System Identification: A Frequency Domain Approach , author=. 2012 , publisher=

    work page 2012

  24. [24]

    Thirty-seventh Conference on Neural Information Processing Systems , year=

    Model-Based Control with Sparse Neural Dynamics , author=. Thirty-seventh Conference on Neural Information Processing Systems , year=

    work page

  25. [25]

    2022 , publisher=

    Regularized system identification: Learning dynamic models from data , author=. 2022 , publisher=

    work page 2022

  26. [26]

    IEEE Control Systems Magazine , volume=

    Nonlinear system identification: A user-oriented road map , author=. IEEE Control Systems Magazine , volume=. 2019 , publisher=

    work page 2019

  27. [27]

    IEEE Transactions on Automatic Control , volume=

    Specialized interior-point algorithm for stable nonlinear system identification , author=. IEEE Transactions on Automatic Control , volume=. 2018 , publisher=

    work page 2018

  28. [28]

    IEEE Transactions on Automatic Control , volume=

    Convex parameterizations and fidelity bounds for nonlinear identification and reduced-order modelling , author=. IEEE Transactions on Automatic Control , volume=. 2017 , publisher=

    work page 2017

  29. [29]

    Automatica , volume=

    Kernel-based identification with frequency domain side-information , author=. Automatica , volume=. 2023 , publisher=

    work page 2023

  30. [30]

    Journal of Process Control , volume=

    On recurrent neural networks for learning-based control: recent results and ideas for future developments , author=. Journal of Process Control , volume=. 2022 , publisher=

    work page 2022

  31. [31]

    Automatica , volume=

    Neural networks for control systems—a survey , author=. Automatica , volume=. 1992 , publisher=

    work page 1992

  32. [32]

    International Journal of Robust and Nonlinear Control , volume=

    Learning model predictive control with long short-term memory networks , author=. International Journal of Robust and Nonlinear Control , volume=. 2021 , publisher=

    work page 2021

  33. [33]

    IEE Proceedings-Control Theory and Applications , volume=

    Dynamic recurrent neural network for system identification and control , author=. IEE Proceedings-Control Theory and Applications , volume=. 1995 , publisher=

    work page 1995

  34. [34]

    Artificial Neural Networks--ICANN 2006: 16th International Conference, Athens, Greece, September 10-14, 2006

    Recurrent neural networks are universal approximators , author=. Artificial Neural Networks--ICANN 2006: 16th International Conference, Athens, Greece, September 10-14, 2006. Proceedings, Part I 16 , pages=. 2006 , organization=

    work page 2006

  35. [35]

    2005 , school=

    Linear Models of Nonlinear Systems , author=. 2005 , school=

    work page 2005

  36. [36]

    Training robust neural networks using

    Pauli, Patricia and Koch, Anne and Berberich, Julian and Kohler, Paul and Allg. Training robust neural networks using. IEEE Control Systems Letters , volume=. 2021 , publisher=

    work page 2021

  37. [37]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Seq2sick: Evaluating the robustness of sequence-to-sequence models with adversarial examples , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

    work page

  38. [38]

    Advances in Neural Information Processing Systems , volume=

    Deep equilibrium models , author=. Advances in Neural Information Processing Systems , volume=

    work page

  39. [39]

    Advances in neural information processing systems , volume=

    Monotone operator equilibrium networks , author=. Advances in neural information processing systems , volume=

    work page

  40. [40]

    2021 60th IEEE Conference on Decision and Control (CDC) , pages=

    Linear systems with neural network nonlinearities: Improved stability analysis via acausal Zames-Falb multipliers , author=. 2021 60th IEEE Conference on Decision and Control (CDC) , pages=. 2021 , organization=

    work page 2021

  41. [41]

    International Journal of Control , volume=

    A shift in paradigm for system identification , author=. International Journal of Control , volume=. 2020 , publisher=

    work page 2020

  42. [42]

    IEEE Transactions on Automatic Control , volume=

    Some new rearrangement inequalities having application in stability analysis , author=. IEEE Transactions on Automatic Control , volume=. 1968 , publisher=

    work page 1968

  43. [43]

    Archive for rational mechanics and analysis , volume=

    Dissipative dynamical systems part I: General theory , author=. Archive for rational mechanics and analysis , volume=. 1972 , publisher=

    work page 1972

  44. [44]

    Trends in control: a European perspective , pages=

    Mixed H_2 / H_ Control , author=. Trends in control: a European perspective , pages=. 1995 , publisher=

    work page 1995

  45. [45]

    1998 , publisher=

    Masubuchi, Izumi and Ohara, Atsumi and Suda, Nobuhide , journal=. 1998 , publisher=

    work page 1998

  46. [46]

    Archive for rational mechanics and analysis , volume=

    Dissipative dynamical systems part II: Linear systems with quadratic supply rates , author=. Archive for rational mechanics and analysis , volume=. 1972 , publisher=

    work page 1972

  47. [47]

    2011 , publisher=

    Handbook of marine craft hydrodynamics and motion control , author=. 2011 , publisher=

    work page 2011

  48. [48]

    , author=

    Measurements of wind-wave growth and swell decay during the Joint North Sea Wave Project (JONSWAP). , author=. Ergaenzungsheft zur Deutschen Hydrographischen Zeitschrift, Reihe A , year=

    work page

  49. [49]

    Proceedings of the 7th IFAC Conference on Manoeuvring and Control of Marine Craft , pages=

    A 4-dof simulink model of a coastal patrol vessel for manoeuvring in waves , author=. Proceedings of the 7th IFAC Conference on Manoeuvring and Control of Marine Craft , pages=. 2006 , organization=

    work page 2006

  50. [50]

    The Royal Institution of Naval Architects , volume=

    Wind resistance of merchant ships , author=. The Royal Institution of Naval Architects , volume=

    work page

  51. [51]

    2021 European Control Conference (ECC) , pages=

    Improved initialization of state-space artificial neural networks , author=. 2021 European Control Conference (ECC) , pages=. 2021 , organization=

    work page 2021

  52. [52]

    2016 , publisher=

    Networks of dissipative systems: compositional certification of stability, performance, and safety , author=. 2016 , publisher=

    work page 2016

  53. [53]

    IEEE transactions on automatic control , volume=

    System analysis via integral quadratic constraints , author=. IEEE transactions on automatic control , volume=. 1997 , publisher=

    work page 1997

  54. [54]

    International Journal of Robust and Nonlinear Control , volume=

    IQC-synthesis with general dynamic multipliers , author=. International Journal of Robust and Nonlinear Control , volume=. 2014 , publisher=

    work page 2014

  55. [55]

    IEEE Control Systems Letters , volume=

    One-shot verification of dissipativity properties from input--output data , author=. IEEE Control Systems Letters , volume=. 2019 , publisher=

    work page 2019

  56. [56]

    IEEE Transactions on Automatic Control , volume=

    Provably robust verification of dissipativity properties from data , author=. IEEE Transactions on Automatic Control , volume=. 2021 , publisher=

    work page 2021

  57. [57]

    International Conference on Machine Learning , pages=

    Learning globally smooth functions on manifolds , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    work page 2023

  58. [58]

    Baier, Alexandra and Aspandi, Decky and Staab, Steffen , booktitle=

    work page

  59. [59]

    F-16 Aircraft Benchmark Based on Ground Vibration Test Data , publisher =

    Noël, Jean-Philippe and Schoukens, Maarten , keywords =. F-16 Aircraft Benchmark Based on Ground Vibration Test Data , publisher =. 2020 , copyright =. doi:10.4121/12954911.v1 , url =

    work page doi:10.4121/12954911.v1 2020

  60. [60]

    Frequency Domain Gaussian Process Models for H\^

    Devonport, Alex and Seiler, Peter and Arcak, Murat , booktitle=. Frequency Domain Gaussian Process Models for H\^. 2023 , organization=

    work page 2023

  61. [61]

    IEEE Control Systems Letters , volume=

    Learning Dissipative Neural Dynamical Systems , author=. IEEE Control Systems Letters , volume=. 2023 , publisher=

    work page 2023

  62. [62]

    AIAA Journal , volume=

    Data-driven aerospace engineering: reframing the industry with machine learning , author=. AIAA Journal , volume=. 2021 , publisher=

    work page 2021

  63. [63]

    Nature communications , volume=

    Deep learning for universal linear embeddings of nonlinear dynamics , author=. Nature communications , volume=. 2018 , publisher=

    work page 2018

  64. [64]

    2022 , publisher=

    Knowledge guided machine learning: Accelerating discovery using scientific knowledge and data , author=. 2022 , publisher=

    work page 2022

  65. [65]

    Automatica , volume=

    Deep subspace encoders for nonlinear system identification , author=. Automatica , volume=. 2023 , publisher=

    work page 2023

  66. [66]

    International Conference on Machine Learning , pages=

    Direct parameterization of lipschitz-bounded deep networks , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    work page 2023

  67. [67]

    Nature Reviews Physics , volume=

    Physics-informed machine learning , author=. Nature Reviews Physics , volume=. 2021 , publisher=

    work page 2021

  68. [68]

    Hysteretic Benchmark with a Dynamic Nonlinearity , publisher =

    Noël, Jean-Philippe and Schoukens, Maarten , keywords =. Hysteretic Benchmark with a Dynamic Nonlinearity , publisher =. 2020 , copyright =. doi:10.4121/12967592.v1 , url =

    work page doi:10.4121/12967592.v1 2020

  69. [69]

    International journal of robust and nonlinear control , volume=

    A linear matrix inequality approach to H _ control , author=. International journal of robust and nonlinear control , volume=. 1994 , publisher=

    work page 1994

  70. [70]

    International journal of control , volume=

    Extended H 2 and H norm characterizations and controller parametrizations for discrete-time systems , author=. International journal of control , volume=. 2002 , publisher=

    work page 2002

  71. [71]

    European Journal of Control , volume=

    Continuous-time system identification with neural networks: Model structures and fitting criteria , author=. European Journal of Control , volume=. 2021 , publisher=

    work page 2021

  72. [72]

    2020 IEEE 14th International Conference on Application of Information and Communication Technologies (AICT) , pages=

    Model structures and fitting criteria for system identification with neural networks , author=. 2020 IEEE 14th International Conference on Application of Information and Communication Technologies (AICT) , pages=. 2020 , organization=

    work page 2020

  73. [73]

    Computationally efficient identification of continuous-time

    Shakib, Mohammad Fahim and Pogromsky, Alexander Yu and Pavlov, Alexey and van de Wouw, Nathan , journal=. Computationally efficient identification of continuous-time. 2022 , publisher=

    work page 2022

  74. [74]

    2023 62nd IEEE Conference on Decision and Control (CDC) , pages=

    Unconstrained parametrization of dissipative and contracting neural ordinary differential equations , author=. 2023 62nd IEEE Conference on Decision and Control (CDC) , pages=. 2023 , organization=

    work page 2023

  75. [75]

    Linear Algebra and its Applications , volume=

    Structured finite-dimensional controller design by convex optimization , author=. Linear Algebra and its Applications , volume=. 2002 , publisher=

    work page 2002

  76. [76]

    Proceedings of the 39th IEEE Conference on Decision and Control (Cat

    Design of structured controllers with applications , author=. Proceedings of the 39th IEEE Conference on Decision and Control (Cat. No. 00CH37187) , volume=. 2000 , organization=

    work page 2000

  77. [77]

    arXiv preprint arXiv:2404.07373 , year=

    Synthesizing Neural Network Controllers with Closed-Loop Dissipativity Guarantees , author=. arXiv preprint arXiv:2404.07373 , year=

    work page arXiv

  78. [78]

    Workshop on nonlinear system identification benchmarks , pages=

    Cascaded tanks benchmark combining soft and hard nonlinearities , author=. Workshop on nonlinear system identification benchmarks , pages=. 2016 , organization=

    work page 2016

  79. [79]

    Antiwindup design with guaranteed regions of stability: an

    Da Silva, JM Gomes and Tarbouriech, Sophie , journal=. Antiwindup design with guaranteed regions of stability: an. 2005 , publisher=

    work page 2005

  80. [80]

    2011 , publisher=

    Stability and stabilization of linear systems with saturating actuators , author=. 2011 , publisher=

    work page 2011

Showing first 80 references.