pith. sign in

arxiv: 2508.07876 · v2 · pith:RN5QCZU7new · submitted 2025-08-11 · 📊 stat.ML · cs.LG· math.DS· math.ST· stat.TH

Stochastic dynamics learning with state-space systems

Juan-Pablo Ortega , Florian Rossmannek This is my paper

Pith reviewed 2026-05-18 23:55 UTC · model grok-4.3

classification 📊 stat.ML cs.LGmath.DSmath.STstat.TH
keywords reservoir computingfading memoryecho state propertystate-space systemsstochastic dynamicsattractor dynamicstime series learning
0
0 comments X

The pith

Fading memory and solution stability hold generically in state-space systems even without the echo state property.

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

The paper establishes that reservoir computing models based on state-space systems display fading memory and stable solutions as typical features across both deterministic and stochastic cases. This occurs without requiring the strict echo state property or contractive conditions that were previously emphasized. A distributional perspective on attractor dynamics in probability space is introduced for the stochastic regime to create a consistent theory. The approach extends prior results on non-autonomous dynamical systems to unify causality, stability, and memory concepts. These findings support reliable generative modeling of temporal data under broader conditions.

Core claim

State-space systems exhibit fading memory and solution stability generically, even in the absence of the echo state property, with a distributional perspective on attractor dynamics providing a coherent theory for the stochastic setting that unifies deterministic and stochastic treatments.

What carries the argument

State-space systems as the central model class, where generic properties are derived by extending results on non-autonomous dynamical systems and applying attractor dynamics on the space of probability distributions for the stochastic case.

If this is right

  • Reservoir computing models can succeed empirically without enforcing strict contractivity conditions.
  • A unified theory covers both deterministic and stochastic regimes for time series learning.
  • Insights into causality, stability, and memory become available for generative modeling of temporal data.
  • The distributional view of attractors supports coherent analysis in stochastic echo state settings.

Where Pith is reading between the lines

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

  • Design of reservoir computing systems could prioritize generic stability over contractivity constraints.
  • Similar distributional attractor ideas might apply to other dynamical learning models beyond state-space systems.
  • Empirical tests could verify stability in non-contractive reservoirs using specific time series benchmarks.

Load-bearing premise

The state-space systems framework permits generic properties to be established by extending prior results on non-autonomous dynamical systems.

What would settle it

A concrete counterexample state-space system without the echo state property in which fading memory or solution stability fails to hold generically would falsify the claim.

Figures

Figures reproduced from arXiv: 2508.07876 by Florian Rossmannek, Juan-Pablo Ortega.

Figure 1
Figure 1. Figure 1: Commutative diagrams depicting the relations in Section [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
read the original abstract

This work advances the theoretical foundations of reservoir computing (RC) by providing a unified treatment of fading memory and the echo state property (ESP) in both deterministic and stochastic settings. We investigate state-space systems, a central model class in time series learning, and establish that fading memory and solution stability hold generically -- even in the absence of the ESP -- offering a robust explanation for the empirical success of RC models without strict contractivity conditions. In the stochastic case, we critically assess stochastic echo states, proposing a novel distributional perspective rooted in attractor dynamics on the space of probability distributions, which leads to a rich and coherent theory. Our results extend and generalize previous work on non-autonomous dynamical systems, offering new insights into causality, stability, and memory in RC models. This lays the groundwork for reliable generative modeling of temporal data in both deterministic and stochastic regimes.

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

2 major / 2 minor

Summary. The manuscript advances theoretical foundations for reservoir computing by providing a unified treatment of fading memory and the echo state property (ESP) for state-space systems in both deterministic and stochastic settings. It claims that fading memory and solution stability hold generically even without the ESP, by extending prior results on non-autonomous dynamical systems, and introduces a distributional perspective on attractors in the space of probability measures to handle the stochastic case, thereby explaining the empirical success of RC models without strict contractivity.

Significance. If the extensions and generic claims are rigorously supported, the work would offer a more general theoretical account of stability and memory in RC models, relaxing reliance on contractivity and enabling broader use in stochastic time-series learning and generative modeling.

major comments (2)
  1. [stochastic case / distributional attractor perspective] The extension of deterministic fixed-point and attractor arguments to the stochastic case via distributional dynamics on probability measures (as described in the abstract and the stochastic section) implicitly requires that the induced map on measures preserves contraction or compactness under weak/Wasserstein topology. This may fail without additional regularity on the noise (e.g., pathwise continuity or Lipschitz/dissipativity conditions on the vector field), particularly in infinite-dimensional state spaces; the manuscript must state and verify these conditions explicitly, as they are load-bearing for the generic holding claim without ESP.
  2. [main results on generic properties / extension of prior non-autonomous results] The central claim that fading memory and solution stability hold generically without the ESP, via extension of non-autonomous dynamical systems results, requires explicit identification of the regularity assumptions that carry over when the driving signal is replaced by a stochastic process. Without this, it is unclear whether the generic property is proven or merely conjectured from the deterministic case.
minor comments (2)
  1. [Abstract] The abstract states that the work 'critically assess[es] stochastic echo states' but does not preview the specific conclusions of that assessment; a brief statement of the outcome would improve clarity.
  2. [preliminaries / notation] Notation for the state-space system and the map on measures should be introduced with a short reminder of the relevant prior definitions to aid readers unfamiliar with the referenced non-autonomous systems literature.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments, which help clarify the presentation of our results on stochastic extensions and generic properties. We address each major comment below, indicating the revisions planned for the next version of the manuscript.

read point-by-point responses

  1. Referee: The extension of deterministic fixed-point and attractor arguments to the stochastic case via distributional dynamics on probability measures (as described in the abstract and the stochastic section) implicitly requires that the induced map on measures preserves contraction or compactness under weak/Wasserstein topology. This may fail without additional regularity on the noise (e.g., pathwise continuity or Lipschitz/dissipativity conditions on the vector field), particularly in infinite-dimensional state spaces; the manuscript must state and verify these conditions explicitly, as they are load-bearing for the generic holding claim without ESP.

    Authors: We agree that the distributional perspective requires explicit regularity conditions to ensure the induced map on probability measures is well-defined and preserves the relevant topological properties under the Wasserstein metric. The manuscript implicitly relies on the noise being a continuous process with Lipschitz and dissipative vector fields, but these were not stated with sufficient detail. In the revision we will add a dedicated paragraph in the stochastic section that lists the precise assumptions (pathwise continuity of the driving noise, uniform Lipschitz continuity of the state map, and dissipativity to guarantee tightness and compactness in the space of measures). We will also include a short verification that these conditions suffice for the generic fading-memory result to hold without the ESP, covering both finite- and infinite-dimensional state spaces via appropriate choice of the underlying Banach space. revision: yes

  2. Referee: The central claim that fading memory and solution stability hold generically without the ESP, via extension of non-autonomous dynamical systems results, requires explicit identification of the regularity assumptions that carry over when the driving signal is replaced by a stochastic process. Without this, it is unclear whether the generic property is proven or merely conjectured from the deterministic case.

    Authors: The extension treats the stochastic input as a random non-autonomous driving signal whose sample paths satisfy the same regularity hypotheses used in the deterministic theory (almost-sure continuity and bounded variation on compact intervals). Under these hypotheses the topological arguments for genericity carry over directly, so the property is proven rather than conjectured. To remove any ambiguity we will insert an explicit remark immediately after the statement of the main generic theorem that enumerates the carried-over regularity conditions on the stochastic process and briefly sketches why the deterministic proof applies pathwise. This makes the logical status of the claim transparent. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation extends prior independent results on non-autonomous systems

full rationale

The paper's central claims on generic fading memory and solution stability (even without ESP) are presented as extensions of prior results on non-autonomous dynamical systems, with a new distributional perspective for the stochastic case. No load-bearing step reduces by construction to fitted inputs, self-definitions, or unverified self-citations; the arguments rely on mathematical properties of state-space systems and attractor dynamics that are externally falsifiable and not tautological. The derivation remains self-contained against the stated modeling framework without circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the paper relies on standard concepts from dynamical systems theory without introducing new free parameters or invented entities.

axioms (1)
  • standard math Standard results from the theory of non-autonomous dynamical systems apply to state-space models in both deterministic and stochastic regimes
    Invoked when extending previous work to establish generic fading memory and stability.

pith-pipeline@v0.9.0 · 5677 in / 1170 out tokens · 47787 ms · 2026-05-18T23:55:34.580225+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.

Forward citations

Cited by 2 Pith papers

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

  1. Echoes of the Past: A Unified Perspective on Fading memory and Echo States

    stat.ML 2025-08 unverdicted novelty 7.0

    The paper unifies fading memory, echo states, and related memory notions in RNNs via new equivalences, implications, and alternative proofs.

  2. Learning the climate of dynamical systems with state-space systems

    nlin.CD 2025-12 unverdicted novelty 5.0

    A C1-close state-space proxy can keep forecasted distributions close to the true long-term distribution for structurally stable mixing dynamical systems.

Reference graph

Works this paper leans on

84 extracted references · 84 canonical work pages · cited by 2 Pith papers · 2 internal anchors

  1. [1]

    Analyzing Patient Trajectories With Artificial Intelligence.J Med Internet Res 23, 12 (Dec 2021), e29812

    Allam, A., Feuerriegel, S., Rebhan, M., and Krauthammer, M. Analyzing Patient Trajectories With Artificial Intelligence.J Med Internet Res 23, 12 (Dec 2021), e29812

    work page 2021

  2. [2]

    M., Dale, R., King, J

    Amigó, J. M., Dale, R., King, J. C., and Lehnertz, K. Generalized synchronization in the presence of dynamical noise and its detection via recurrent neural networks.Chaos: An Interdisciplinary Journal of Nonlinear Science 34, 12 (Dec 2024), 123156

    work page 2024

  3. [3]

    C., V an der Sande, G., Danckaert, J., Massar, S., Dambre, J., Schrauwen, B., Mirasso, C

    Appeltant, L., Soriano, M. C., V an der Sande, G., Danckaert, J., Massar, S., Dambre, J., Schrauwen, B., Mirasso, C. R., and Fischer, I. Information processing using a single dynamical node as complex system. Nature Communications 2, 1 (2011), 468

    work page 2011

  4. [4]

    R., and Ott, E

    Arcomano, T., Szunyogh, I., Wikner, A., Pathak, J., Hunt, B. R., and Ott, E. A hybrid approach to atmospheric modeling that combines machine learning with a physics-based numerical model.Journal of Advances in Modeling Earth Systems 14, 3 (2022), e2021MS002712

    work page 2022

  5. [5]

    Random Dynamical Systems, 1 ed

    Arnold, L. Random Dynamical Systems, 1 ed. Springer Monographs in Mathematics. Springer Berlin, Heidel- berg, 1998

    work page 1998

  6. [6]

    Set-Valued Analysis, 1 ed., vol

    Aubin, J.-P., and Frankowska, H. Set-Valued Analysis, 1 ed., vol. 234 ofModern Birkhäuser Classics. Birkhäuser Boston, 2009

    work page 2009

  7. [7]

    V., and Pilyugin, S

    Babin, A. V., and Pilyugin, S. Y. Continuous dependence of attractors on the shape of domain.Journal of Mathematical Sciences 87, 2 (Nov 1997), 3304–3310. transl. from Zap. Nauchn. Sem. POMI, 221 (1995), p. 58–66

    work page 1997

  8. [8]

    Reser- voir computing for macroeconomic forecasting with mixed-frequency data.International Journal of Forecasting 40, 3 (2024), 1206–1237

    Ballarin, G., Dellaportas, P., Grigoryev a, L., Hirt, M., v an Huellen, S., and Ortega, J.-P. Reser- voir computing for macroeconomic forecasting with mixed-frequency data.International Journal of Forecasting 40, 3 (2024), 1206–1237

    work page 2024

  9. [9]

    Learning Theory for Dynamical Systems

    Berry, T., and Das, S. Learning Theory for Dynamical Systems. SIAM Journal on Applied Dynamical Systems 22, 3 (2023), 2082–2122

    work page 2023

  10. [10]

    Bogachev, V. I. Measure Theory, 1 ed. Springer-Verlag, Berlin Heidelberg, 2007. 27 Ortega and Rossmannek

    work page 2007

  11. [11]

    Power Hungry: How AI Will Drive Energy Demand.IMF Working Papers 2025, 81 (4 2025), 32

    Bogmans, C., Gomez-Gonzalez, P., Ganpurev, G., Melina, G., Pescatori, A., and Thube, S. Power Hungry: How AI Will Drive Energy Demand.IMF Working Papers 2025, 81 (4 2025), 32

    work page 2025

  12. [12]

    General Topology, 1 ed

    Bourbaki, N. General Topology, 1 ed. Springer Berlin, Heidelberg, 1998

    work page 1998

  13. [13]

    Fading memory and the problem of approximating nonlinear operators with Volterra series

    Boyd, S., and Chua, L. Fading memory and the problem of approximating nonlinear operators with Volterra series. IEEE Transactions on Circuits and Systems 32, 11 (1985), 1150–1161

    work page 1985

  14. [14]

    Generating Financial Markets With Signatures

    Buehler, H., Hor v ath, B., Lyons, T., Perez Arribas, I., and Wood, B. Generating Financial Markets With Signatures. SSRN:3657366 (2020)

    work page 2020

  15. [15]

    A tighter bound for the echo state property.IEEE Transactions on Neural Networks 17, 3 (2006), 820–824

    Buehner, M., and Young, P. A tighter bound for the echo state property.IEEE Transactions on Neural Networks 17, 3 (2006), 820–824

    work page 2006

  16. [16]

    The echo index and multistability in input-driven recurrent neural networks.Physica D: Nonlinear Phenomena 412(2020), 132609

    Ceni, A., Ashwin, P., Livi, L., and Postlethw aite, C. The echo index and multistability in input-driven recurrent neural networks.Physica D: Nonlinear Phenomena 412(2020), 132609

    work page 2020

  17. [17]

    2023 , note =

    Cheridito, P., and Eckstein, S. Optimal transport and Wasserstein distances for causal models. arXiv:2303.14085v2 (2023)

    work page arXiv 2023

  18. [18]

    S., and Teicher, H

    Chow, Y. S., and Teicher, H. Probability Theory: Independence, Interchangeability, Martingales, 3 ed. Springer Texts in Statistics. Springer New York, 1997

    work page 1997

  19. [19]

    Chua, L., and Green, D. A qualitative analysis of the behavior of dynamic nonlinear networks: Steady-state solutions of nonautonomous networks.IEEE Transactions on Circuits and Systems 23, 9 (1976), 530–550

    work page 1976

  20. [20]

    Attractors for random dynamical systems.Probability Theory and Related Fields 100, 3 (Sep 1994), 365–393

    Crauel, H., and Flandoli, F. Attractors for random dynamical systems.Probability Theory and Related Fields 100, 3 (Sep 1994), 365–393

    work page 1994

  21. [21]

    Desheng, L., and Kloeden, P. E. Equi-attraction and the continuous dependence of attractors on parameters. Glasgow Mathematical Journal 46, 1 (2004), 131–141

    work page 2004

  22. [22]

    General Topology, vol

    Engelking, R. General Topology, vol. 6 ofSigma Series in Pure Mathematics. Heldermann Verlag Berlin, 1989

    work page 1989

  23. [23]

    Eroglu, D., Lamb, J. S. W., and Pereira, T. Synchronisation of chaos and its applications.Contemporary Physics 58, 3 (2017), 207–243

    work page 2017

  24. [24]

    Pattern Recognition in a Bucket

    Fernando, C., and Sojakka, S. Pattern Recognition in a Bucket. InAdvances in Artificial Life(2003), W. Banzhaf, J. Ziegler, T. Christaller, P. Dittrich, and J. T. Kim, Eds., Springer Berlin Heidelberg, pp. 588–597

    work page 2003

  25. [25]

    GARCH models, 2 ed

    Francq, C., and Zakoian, J.-M. GARCH models, 2 ed. John Wiley & Sons, 2019

    work page 2019

  26. [26]

    Approximation bounds for random neural networks and reservoir systems.The Annals of Applied Probability 33, 1 (2023), 28–69

    Gonon, L., Grigoryev a, L., and Ortega, J.-P. Approximation bounds for random neural networks and reservoir systems.The Annals of Applied Probability 33, 1 (2023), 28–69

    work page 2023

  27. [27]

    Fading memory echo state networks are universal.Neural Networks 138(2021), 10–13

    Gonon, L., and Ortega, J.-P. Fading memory echo state networks are universal.Neural Networks 138(2021), 10–13

    work page 2021

  28. [28]

    Generating Sequences With Recurrent Neural Networks

    Gra ves, A. Generating Sequences With Recurrent Neural Networks.arXiv:1308.0850v5 (2014)

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  29. [29]

    G., and Ortega, J.-P

    Grigoryev a, L., Hart, A. G., and Ortega, J.-P. Chaos on compact manifolds: Differentiable synchro- nizations beyond the Takens theorem.Physical Review E - Statistical Physics, Plasmas, Fluids, and Related Interdisciplinary Topics 103(2021), 062204

    work page 2021

  30. [30]

    G., and Ortega, J.-P

    Grigoryev a, L., Hart, A. G., and Ortega, J.-P. Learning strange attractors with reservoir systems. Nonlinearity 36 (2023), 4674–4708

    work page 2023

  31. [31]

    Echo state networks are universal.Neural Networks 108(2018), 495–508

    Grigoryev a, L., and Ortega, J.-P. Echo state networks are universal.Neural Networks 108(2018), 495–508

    work page 2018

  32. [32]

    Grigoryev a, L., and Ortega, J.-P. Universal discrete-time reservoir computers with stochastic inputs and linear readouts using non-homogeneous state-affine systems.Journal of Machine Learning Research 19, 24 (2018), 1–40

    work page 2018

  33. [33]

    Differentiable reservoir computing.Journal of Machine Learning Research 20, 179 (2019), 1–62

    Grigoryev a, L., and Ortega, J.-P. Differentiable reservoir computing.Journal of Machine Learning Research 20, 179 (2019), 1–62

    work page 2019

  34. [34]

    Dimension reduction in recurrent networks by canonicalization.Journal of Geometric Mechanics 13, 4 (2021), 647–677

    Grigoryev a, L., and Ortega, J.-P. Dimension reduction in recurrent networks by canonicalization.Journal of Geometric Mechanics 13, 4 (2021), 647–677

    work page 2021

  35. [35]

    Mamba: Linear-Time Sequence Modeling with Selective State Spaces

    Gu, A., and Dao, T. Mamba: Linear-Time Sequence Modeling with Selective State Spaces.arXiv:2312.00752v2 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  36. [36]

    Recurrent Neural Networks for Time Series Forecasting: Current status and future directions.International Journal of Forecasting 37, 1 (2021), 388–427

    Hew amalage, H., Bergmeir, C., and Bandara, K. Recurrent Neural Networks for Time Series Forecasting: Current status and future directions.International Journal of Forecasting 37, 1 (2021), 388–427

    work page 2021

  37. [37]

    T., Olson, E

    Hoang, L. T., Olson, E. J., and Robinson, J. C. On the continuity of global attractors.Proceedings of the American Mathematical Society 143, 10 (2015), 4389–4395

    work page 2015

  38. [38]

    Sur un type de convergence intermediaire entre la convergence en loi et la convergence en probabilite

    Jacod, J., and Memin, J. Sur un type de convergence intermediaire entre la convergence en loi et la convergence en probabilite. InSéminaire de Probabilités XV 1979/80(1981), J. Azéma and M. Yor, Eds., vol. 850 ofLecture Notes in Mathematics, Springer Berlin Heidelberg, pp. 529–546

    work page 1979

  39. [39]

    echo state

    Jaeger, H. The “echo state” approach to analysing and training recurrent neural networks – with an Erratum note. Tech. Rep. GMD Report 148, German National Research Center for Information Technology, 2010. 28 Stochastic dynamics learning with state-space systems

    work page 2010

  40. [40]

    Harnessing nonlinearity: Predicting chaotic systems and saving energy in wireless communication

    Jaeger, H., and Haas, H. Harnessing nonlinearity: Predicting chaotic systems and saving energy in wireless communication. Science 304, 5667 (2004), 78–80

    work page 2004

  41. [41]

    A Brief Survey on the Approximation Theory for Sequence Modelling

    Jiang, H., Li, Q., Li, Z., and W ang, S. A Brief Survey on the Approximation Theory for Sequence Modelling. Journal of Machine Learning 2, 1 (2023), 1–30

    work page 2023

  42. [42]

    Preventing the Immense Increase in the Life-Cycle Energy and Carbon Footprints of LLM-Powered Intelligent Chatbots.Engineering 40 (2024), 202–210

    Jiang, P., Sonne, C., Li, W., You, F., and You, S. Preventing the Immense Increase in the Life-Cycle Energy and Carbon Footprints of LLM-Powered Intelligent Chatbots.Engineering 40 (2024), 202–210

    work page 2024

  43. [43]

    Foundations of Modern Probability, 3 ed., vol

    Kallenberg, O. Foundations of Modern Probability, 3 ed., vol. 99 ofProbability Theory and Stochastic Modelling. Springer Nature Switzerland, 2021

    work page 2021

  44. [44]

    Kalman, R. E. A new approach to linear filtering and prediction problems.Journal of Basic Engineering 82, 1 (1960), 35–45

    work page 1960

  45. [45]

    S., Maciejowski, J., and Chopin, N

    Kantas, N., Doucet, A., Singh, S. S., Maciejowski, J., and Chopin, N. On Particle Methods for Parameter Estimation in State-Space Models.Statistical Science 30, 3 (2015), 328 – 351

    work page 2015

  46. [46]

    E., Pötzsche, C., and Rasmussen, M

    Kloeden, P. E., Pötzsche, C., and Rasmussen, M. Limitations of pullback attractors for processes.Journal of Difference Equations and Applications 18, 4 (2012), 693–701

    work page 2012

  47. [47]

    E., Pötzsche, C., and Rasmussen, M

    Kloeden, P. E., Pötzsche, C., and Rasmussen, M. Discrete-Time Nonautonomous Dynamical Systems. In Stability and Bifurcation Theory for Non-Autonomous Differential Equations, R. Johnson and M. Pera, Eds., vol. 2065 ofLecture Notes in Mathematics. Springer Berlin, Heidelberg, 2013, pp. 35–102

    work page 2065

  48. [48]

    E., and Rasmussen, M

    Kloeden, P. E., and Rasmussen, M. Nonautonomous Dynamical Systems, vol. 176 ofMathematical Surveys and Monographs. American Mathematical Society, 2011

    work page 2011

  49. [49]

    Generalized Synchronization, Predictability, and Equivalence of Unidirectionally Coupled Dynamical Systems.Phys

    Kocarev, L., and Parlitz, U. Generalized Synchronization, Predictability, and Equivalence of Unidirectionally Coupled Dynamical Systems.Phys. Rev. Lett. 76(Mar 1996), 1816–1819

    work page 1996

  50. [50]

    P., Exarchos, K

    Kourou, K., Exarchos, T. P., Exarchos, K. P., Karamouzis, M. V., and Fotiadis, D. I. Machine learning applications in cancer prognosis and prediction.Computational and Structural Biotechnology Journal 13 (2015), 8–17

    work page 2015

  51. [51]

    A review of unsupervised feature learning and deep learning for time-series modeling.Pattern Recognition Letters 42(2014), 11–24

    Längkvist, M., Karlsson, L., and Loutfi, A. A review of unsupervised feature learning and deep learning for time-series modeling.Pattern Recognition Letters 42(2014), 11–24

    work page 2014

  52. [52]

    C., Brunner, D., Appeltant, L., Gutierrez, J

    Larger, L., Soriano, M. C., Brunner, D., Appeltant, L., Gutierrez, J. M., Pesquera, L., Mirasso, C. R., and Fischer, I. Photonic information processing beyond Turing: an optoelectronic implementation of reservoir computing.Opt. Express 20, 3 (Jan 2012), 3241–3249

    work page 2012

  53. [53]

    Graphical Models, vol

    Lauritzen, S. Graphical Models, vol. 17 ofOxford Statistical Science Series. Clarendon Press, Oxford, 1996

    work page 1996

  54. [54]

    Total Variation Convergence Preserves Conditional Independence.arXiv:2401.06177v1 (2024)

    Lauritzen, S. Total Variation Convergence Preserves Conditional Independence.arXiv:2401.06177v1 (2024)

    work page arXiv 2024

  55. [55]

    Deep learning

    LeCun, Y., Bengio, Y., and Hinton, G. Deep learning. Nature 521, 7553 (2015), 436–444

    work page 2015

  56. [56]

    Time-series forecasting with deep learning: a survey.Philosophical Transactions of the Royal Society A 379, 2194 (2021), 20200209

    Lim, B., and Zohren, S. Time-series forecasting with deep learning: a survey.Philosophical Transactions of the Royal Society A 379, 2194 (2021), 20200209

    work page 2021

  57. [57]

    Ancestor Sampling for Particle Gibbs

    Lindsten, F., Schön, T., and Jordan, M. Ancestor Sampling for Particle Gibbs. InAdvances in Neural Information Processing Systems(2012), F. Pereira, C. Burges, L. Bottou, and K. Weinberger, Eds., vol. 25, Curran Associates, Inc., pp. 1–9

    work page 2012

  58. [58]

    Machine learning for synthetic data generation: a review.arXiv preprint arXiv:2302.04062,

    Lu, Y., Chen, L., Zhang, Y., Shen, M., W ang, H., W ang, X., v an Rechem, C., Fu, T., and Wei, W. Machine Learning for Synthetic Data Generation: A Review.arXiv:2302.04062v10 (2025)

    work page arXiv 2025

  59. [59]

    R., and Ott, E

    Lu, Z., Hunt, B. R., and Ott, E. Attractor reconstruction by machine learning.Chaos 28, 6 (2018)

    work page 2018

  60. [60]

    Liquid State Machines: Motivation, Theory, and Applications

    Maass, W. Liquid State Machines: Motivation, Theory, and Applications. In Computability In Context: Computation And Logic In The Real World, S. B. Cooper and A. Sorbi, Eds. Imperial College Press, 2011, pp. 275–296

    work page 2011

  61. [61]

    Real-Time Computing Without Stable States: A New Framework for Neural Computation Based on Perturbations.Neural Computation 14, 11 (Nov 2002), 2531–2560

    Maass, W., Natschläger, T., and Markram, H. Real-Time Computing Without Stable States: A New Framework for Neural Computation Based on Perturbations.Neural Computation 14, 11 (Nov 2002), 2531–2560

    work page 2002

  62. [62]

    Stability and memory-loss go hand-in-hand: three results in dynamics and computation

    Manjunath, G. Stability and memory-loss go hand-in-hand: three results in dynamics and computation. Proceedings of the Royal Society A 476(2020), 20200563

    work page 2020

  63. [63]

    Embedding information onto a dynamical system.Nonlinearity 35, 3 (Jan 2022), 1131

    Manjunath, G. Embedding information onto a dynamical system.Nonlinearity 35, 3 (Jan 2022), 1131

    work page 2022

  64. [64]

    Echo State Property Linked to an Input: Exploring a Fundamental Characteristic of Recurrent Neural Networks.Neural Computation 25, 3 (2013), 671–696

    Manjunath, G., and Jaeger, H. Echo State Property Linked to an Input: Exploring a Fundamental Characteristic of Recurrent Neural Networks.Neural Computation 25, 3 (2013), 671–696

    work page 2013

  65. [65]

    The Dynamics of Random Difference Equations Is Remodeled by Closed Relations

    Manjunath, G., and Jaeger, H. The Dynamics of Random Difference Equations Is Remodeled by Closed Relations. SIAM Journal on Mathematical Analysis 46, 1 (2014), 459–483

    work page 2014

  66. [66]

    Transport in reservoir computing.Physica D: Nonlinear Phenomena 449 (2023), 133744

    Manjunath, G., and Ortega, J.-P. Transport in reservoir computing.Physica D: Nonlinear Phenomena 449 (2023), 133744

    work page 2023

  67. [67]

    Quantum reservoir computing in finite dimensions.Phys

    Martínez-Peña, R., and Ortega, J.-P. Quantum reservoir computing in finite dimensions.Phys. Rev. E 107 (Mar 2023), 035306

    work page 2023

  68. [68]

    On the concept of attractor.Communications in Mathematical Physics 99, 2 (1985), 177–195

    Milnor, J. On the concept of attractor.Communications in Mathematical Physics 99, 2 (1985), 177–195. 29 Ortega and Rossmannek

    work page 1985

  69. [69]

    Physical reservoir computing—an introductory perspective.Japanese Journal of Applied Physics 59, 6 (May 2020), 060501

    Nakajima, K. Physical reservoir computing—an introductory perspective.Japanese Journal of Applied Physics 59, 6 (May 2020), 060501

    work page 2020

  70. [70]

    Measure and Statistical Attractors for Nonautonomous Dynamical Systems

    Oljača, L., Ashwin, P., and Rasmussen, M. Measure and Statistical Attractors for Nonautonomous Dynamical Systems. Journal of Dynamics and Differential Equations 36, 3 (Sep 2024), 2375–2411

    work page 2024

  71. [71]

    Echoes of the past: a unified perspective on fading memory and echo states

    Ortega, J.-P., and Rossmannek, F. Echoes of the past: a unified perspective on fading memory and echo states

    work page

  72. [72]

    Fading memory and the convolution theorem.IEEE Transactions on Automatic Control(2025), 1–13

    Ortega, J.-P., and Rossmannek, F. Fading memory and the convolution theorem.IEEE Transactions on Automatic Control(2025), 1–13

    work page 2025

  73. [73]

    State-space systems as dynamic generative models.Proceedings of the Royal Society A 481, 2309 (2025), 20240308

    Ortega, J.-P., and Rossmannek, F. State-space systems as dynamic generative models.Proceedings of the Royal Society A 481, 2309 (2025), 20240308

    work page 2025

  74. [74]

    A Strong Version of the Skorohod Representation Theorem.Journal of Theoretical Probability 36, 1 (Mar 2023), 372–389

    Pratelli, L., and Rigo, P. A Strong Version of the Skorohod Representation Theorem.Journal of Theoretical Probability 36, 1 (Mar 2023), 372–389

    work page 2023

  75. [75]

    Bayesian Filtering and Smoothing, 1 ed., vol

    Särkkä, S. Bayesian Filtering and Smoothing, 1 ed., vol. 3 ofInstitute of Mathematical Statistics Textbooks. Cambridge University Press, 2013

    work page 2013

  76. [76]

    Deep learning in neural networks: An overview.Neural Networks 61(2015), 85–117

    Schmidhuber, J. Deep learning in neural networks: An overview.Neural Networks 61(2015), 85–117

    work page 2015

  77. [77]

    Sontag, E. D. Mathematical Control Theory: Deterministic Finite Dimensional Systems, 1 ed. Texts in Applied Mathematics. Springer-Verlag New York, 1990

    work page 1990

  78. [78]

    Modeling financial time-series with generative adversarial networks

    Takahashi, S., Chen, Y., and Tanaka-Ishii, K. Modeling financial time-series with generative adversarial networks. Physica A: Statistical Mechanics and its Applications 527(2019), 121261

    work page 2019

  79. [79]

    Detecting strange attractors in turbulence

    Takens, F. Detecting strange attractors in turbulence. InDynamical Systems and Turbulence, D. Rand and L.-S. Young, Eds., 1 ed., vol. 898 ofLecture Notes in Mathematics. Springer Berlin, Heidelberg, 1981, pp. 366–381

    work page 1981

  80. [80]

    Improving Deep Learning with Generic Data Augmentation

    Taylor, L., and Nitschke, G. Improving Deep Learning with Generic Data Augmentation. In2018 IEEE Symposium Series on Computational Intelligence (SSCI)(2018), pp. 1542–1547

    work page 2018

Showing first 80 references.