Pith. sign in

REVIEW 4 major objections 5 minor 70 references

Model-free Forecasting of Rogue Waves using Reservoir Computing

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A reservoir computer, trained on a single quasi-periodic NLS breather, forecasts rogue waves it has never seen, including higher-order breathers and random seas; adding sparse data assimilation and norm conservation lengthens autonomous…

desk verdict Credible RC demo for NLS rogue waves, but the one-step claims need a persistence baseline and the 'long horizon' is really a few dimensionless time units. read the letter →

arxiv 2506.21918 v2 pith:5GKPPMZC submitted 2025-06-27 cs.CE nlin.PS

Abrari Noor Hasmi , Hadi Susanto This is my paper · ORCID
classification cs.CEnlin.PS MSC 68T0735Q55
keywords RoguewavesNonlinearSchrödingerequationReservoircomputingEchoStateNetworkModulationinstabilityHamiltoniansystemBreatherDataassimilation
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

The paper claims that a reservoir computer—a recurrent network whose fixed random middle layer is never trained—can forecast rogue waves in the focusing nonlinear Schrödinger equation, a Hamiltonian system with no attractor to learn. Trained on a single 70-time-unit stretch of a breather-like recurrence, the network one-step-matches a continuation of that data, a fifth-order maximal breather built by a Darboux transformation, and synthetic JONSWAP ocean waves, all without retraining. Autonomous (feedback) prediction is shorter-lived and tends to underestimate peaks because the reservoir dissipates energy and loses the Hamiltonian structure; the paper's combined fix, sparse data assimilation plus per-step rescaling of the solution norm, restores both amplitude and timing. If the claim holds, a cheap data-driven surrogate trained on numerically cheap quasi-periodic data could give short-term warning of extreme waves when the governing model is unavailable.

What carries the argument

The central object is a parallel Echo State Network: 64 reservoirs of 800 hidden nodes each, whose inputs overlap on neighbouring spatial windows and which share a single output matrix computed by ridge regression. Sharing the output matrix encodes the translation invariance of the NLS and, together with the odd symmetry of tanh, matches the symmetries of the underlying equation. The companion mechanism is a two-part stabilizer for autonomous prediction: partial updates that overwrite the feedback input with true data at sparse intervals, and a per-step rescaling of the predicted field that fixes the global solution norm to its initial value.

What would settle it

Train the identical parallel Echo State Network on a segment of the same Ω≈0.39799 recurrence that stays entirely in the quiescent background and never approaches a homoclinic crossing; if the one-step NRMSE on the fifth-order maximal breather no longer stays small near the peak, the reported generalization depends on the specific training trajectory rather than on the architecture. A direct experimental check is to vary the initial perturbation amplitude A1 of the training run and measure how the prediction horizon and peak underestimation change as the sampled phase space moves away from the homoclinic manifold.

Watch

Extended reading notes

Core claim

A properly tuned parallel Echo State Network, trained only on the quasi-periodic recurrence of an Ω ≈ 0.39799 breather with five unstable modes, predicts the one-step dynamics of three different datasets: the continuation of its own training trajectory, a fifth-order maximal intensity breather generated by a Darboux transformation, and random long-crested ocean waves drawn from a JONSWAP spectrum. One-step normalized errors stay small except near the instability peak, and autonomous forecasts give lead times of about two dimensionless time units ahead of a breather peak in the training continuation and around 0.5 time units for the maximal breather. The authors identify the obstacle for longer autonomous runs: the reservoir is intrinsically dissipative, so the predicted solution loses energy and drifts off the Hamiltonian flow; their remedy rescales the predicted field to conserve the initial norm and intermittently injects ground-truth data, which together capture both the amplitude and the timing of the rogue wave.

Load-bearing premise

The claim rests on one 70-time-unit training trajectory (after a 15-unit burn-in) adequately sampling the phase space of the NLS, including the unstable manifold around homoclinic orbits, so that the same reservoir generalizes without retraining to a maximal breather and to random seas.

Editorial extensions

If this is right

  • Without any retraining, the same reservoir one-step-predicts wave fields whose spatial structure differs from its training data: a fifth-order maximal breather and JONSWAP random seas.
  • Autonomous prediction already yields a lead time of roughly two dimensionless units before a breather peak in the continuation data, and about half a unit before the maximal breather peak.
  • The norm-preserving and data-assimilation corrections, applied together, capture both amplitude and timing of a rogue wave that occurs about 5.64 time units after the forecast starts; applying either correction alone degrades one of the two.
  • Training data that also includes random wave realizations (the RC_Cm model) broadens the sampled phase space and improves autonomous amplitude accuracy, even though the reservoir and input matrices are unchanged.
  • The loss of the Hamiltonian property in autonomous mode is the dominant error source; energy diagnostics (kinetic, potential, and total Hamiltonian) track the divergence of the forecast.

Reading between the lines

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

  • A practical deployment recipe follows implicitly: train the reservoir on cheap breather simulations, then use occasional buoy or gauge measurements as the assimilation updates—the norm rescaling needs no model at all.
  • The phase-space-coverage requirement suggests an active sampling strategy: choose the training window adaptively by maximizing overlap with the unstable manifold, using the local stretching number as a guide, rather than taking an arbitrary 70-unit segment.
  • The same combined correction likely transfers to other conservative wave models (e.g., modified NLS or the Davey-Stewartson system) wherever energy drift dominates the autonomous error.
  • A sharper test of the generalization claim would be to train on a recurrence with different modulation parameters and test on the maximal breather of this paper; the one-step error near the peak should stay comparable if the mechanism is robust.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper trains a parallel echo state network (RC_AB) on a focusing NLS simulation with five unstable modes and tests it on three types of data: a temporal continuation of the training trajectory, a fifth-order maximal intensity breather, and JONSWAP random sea states. Two prediction modes are considered: teacher-forced one-step prediction and autonomous (generative) prediction. The paper reports low one-step NRMSE on all test sets, modest autonomous prediction horizons (roughly 0.5-2 dimensionless time units), and proposes two corrections for autonomous mode: intermittent data assimilation (Section 6, Eq. (12)) and a per-step L2-norm rescaling (Eqs. (13)-(14)). The authors argue that the reservoir generalizes to unseen dynamics and that phase-space coverage of the training data is critical for Hamiltonian systems.

Significance. If the claims are fully supported, the paper would be a useful contribution to reservoir computing for Hamiltonian spatiotemporal systems, where attractor-based training arguments do not apply. The held-out random-realization test in Section 5 is a genuine strength, as is the explicit discussion of phase-space coverage in Section 7. The paper also gives credit for using translation-invariant parallel reservoirs and for attempting invariant-preserving corrections in autonomous mode. However, the central generalization claim is currently supported mainly by one-step NRMSE values that lack a persistence baseline, and the 'relatively long prediction horizon' is not quantified against any reference timescale. These gaps make the strongest form of the abstract's claim premature.

major comments (4)
  1. [Sections 2.3, 3.2, 4, 5 (Figs. 3, 8, 10)] The one-step prediction results are not benchmarked against a persistence baseline. With the reservoir sampling interval Δt_RC = 5×10^-3 (Section 3.2), the carrier phase of the NLS solution advances by only about 0.005 radians per step, and the fastest unstable mode (γ_max = 0.964) changes the envelope by roughly 0.5% per step. A persistence forecast ŷ_{j+1} = y_j, or the phase-rotated version ŷ_{j+1} = e^{iΔt_RC} y_j, would already produce very small NRMSE in smooth regions. Because the readout is trained to map the current state to its successor, low one-step error can reflect approximation of a near-identity map rather than learning of the NLS vector field or homoclinic structure. Please report NRMSE for these baselines alongside Figs. 3, 8, and 10, and discuss whether the RC one-step error is actually lower. The autonomous-mode figures are more discriminating and may survive, but the one-step evidence for 'unseen dynamics' is not conclusive without this baseline.
  2. [Abstract; Sections 3.3, 4, 5] The claim of a 'relatively long prediction horizon' is overstated relative to the reported results. Section 3.3 states that the RC can detect rogue waves approximately two time units before occurrence; Section 4 reports a lead time of 0.5 time units for the maximal intensity breather; Section 5 initiates autonomous prediction 0.13 dimensionless time units before the rogue wave event. Since γ_max^-1 ≈ 1.04 time units, the demonstrated horizons are of order one characteristic instability time, not clearly 'long' in any absolute sense. Please quantify the prediction horizon with explicit PH values at several starting times, compare against persistence-based autonomous prediction, and temper the abstract's 'relatively long prediction horizon' wording accordingly.
  3. [Sections 3.2 and 7] The conclusion that training data must sufficiently sample the phase space is explicitly stated in Section 7, but the paper does not provide evidence that the single 70-time-unit training trajectory from the Ω = 0.39799 breather simulation adequately spans the unstable manifold. This assumption is load-bearing for the generalization claims to the fifth-order maximal breather and to JONSWAP seas. Please support it with a concrete test, for example by training on multiple different time windows or multiple initial perturbations and showing that the prediction horizon and one-step errors are stable, or by measuring some proxy of phase-space coverage (e.g., local dimension or distribution of stretching numbers). Without such evidence, the phase-space coverage interpretation remains an untested hypothesis rather than a demonstrated mechanism.
  4. [Section 6, Eqs. (12)-(14), Figs. 12-13] The combined data-assimilation and norm-rescaling method is claimed to 'significantly extend the prediction horizon,' but no quantitative PH values are reported for the three scenarios (update only, normalization only, combined), and the method uses external ground-truth measurements at prescribed intervals. Please report the autonomous prediction horizon for each scenario and for several starting times t0, and clarify how much of the improvement is attributable to the intermittent injection of true data versus the norm rescaling. Note also that Eq. (14) imposes the initial L2 norm at every step by construction; this is a valid constraint for the NLS, whose L2 norm is conserved, but it is a hard constraint applied externally rather than a learned property, and it does not correct phase or timing errors, as Fig. 12 itself shows for the normalization-only case.
minor comments (5)
  1. [Section 2.3, Eq. (10)] In the definition of e_{j0,jl}, the ground-truth term in the numerator and denominator should be y_{j0+jl}, not y_j, to make the index conventions consistent with the preceding sentence.
  2. [Section 3.3 and Appendix B] The text says γ_max is 'equivalent to the largest Lyapunov exponent in the case of a plane wave solution,' while Appendix B states that the largest Lyapunov exponent is 'not well-defined' due to persistent fluctuations of the stretching number. Please reconcile these statements or clarify the distinction between the local stretching rate and a global Lyapunov exponent.
  3. [Section 5 and Appendix C] The time conversion in Eq. (C.1) uses the parameters T_p = 8 s and H_s = 8 m, but the paper does not state the value of the NLS scaling parameter ǫ used to set ||ψ|| = 1. Please specify this value, as it is needed to reproduce the reported physical times such as 1.32 s.
  4. [Data availability statement] The data availability statement says data are available in the article or supplementary materials, but no repository link, code, or detailed hyperparameter configuration file is provided. For reproducibility of the RC results, please make the training and testing datasets and the implementation available.
  5. [Throughout] There are several typos and formatting artifacts in the manuscript text, including 'Numerical Simulat ion' in the journal header and inconsistent spacing in equations; a careful proofread would improve presentation.

Circularity Check

0 steps flagged · score 1.0 of 10

No load-bearing circularity: the readout is fit to a held-out training segment and the main autonomous results are not forced by construction.

full rationale

The study is a supervised reservoir-computing experiment, not a derivation from first principles. The readout matrix Wo is the only trained parameter, obtained by ridge regression (Eqs. 6-8) on a 70-time-unit segment of one NLS breather simulation; the test sets (continuation, fifth-order maximal breather, JONSWAP realizations) are not used to solve for Wo. RC_Cm adds one random-wave realization to the training data but is evaluated on different realizations, so its assessment remains held out. The Section 6 corrections are explicitly external: data assimilation injects ground-truth states at sparse times (Eq. 12), and the norm rescaling (Eqs. 13-14) uses the L2 norm, which is a conserved quantity of the focusing NLS, not a learned parameter. No uniqueness theorem, self-citation chain, or renamed known result carries the argument. The paper even states the relevant limitation explicitly: 'the training data must sufficiently sample the phase space' (Section 7), conceding that generalization is conditional on phase-space coverage rather than guaranteed by construction. The one-step teacher-forcing results lack a persistence baseline, so low NRMSE may partly reflect the near-identity character of the map at dt_RC=5e-3, but this is a benchmarking gap, not circularity, because the reported errors are measured on unseen data and the autonomous predictions provide independent evidence.

Assumptions & free parameters 11 free parameters · 6 assumptions · 0 invented entities

The central claims depend on the trained readout matrix (a fitted object) and on a set of hand-chosen hyperparameters. Beyond that, the paper relies on standard ML assumptions (echo state property) and on domain assumptions about NLS as ground truth for rogue waves. No new physical entities are introduced. The norm-rescaling correction in Section 6 is an ad hoc fix, which is a notable but explicit added assumption.

free parameters (11)
  • Reservoir size per parallel RC (dS) = 800
    Manually chosen; the paper states further increases did not change results (Section 3.2).
  • Number of parallel reservoirs (M) = 64
    Fixed by spatial grid size and output dimension; hand-picked (Section 3.2).
  • Overlap length (l) = 4 (two physical nodes in real/imag parts)
    Chosen to connect adjacent reservoirs; no sensitivity study (Section 3.2).
  • Input noise variance (sigma) = 0.02
    Selected empirically because it improved autonomous prediction; optimal value unknown (Section 3.2).
  • Tikhonov regularization (beta) = 1e-4
    Set by hand; described as not highly sensitive (Section 3.2).
  • Washout steps (Nw) = 100
    Standard warm-up; fixed (Section 3.2).
  • Sampling interval (Delta t_RC) = 0.005
    Downsampled from 2e-4 simulation step; chosen for training (Section 3.2).
  • Spectral radius of reservoir matrix
    Adjusted during initialization but numerical value not reported (Section 2.2).
  • Input scaling (alpha)
    Entries uniformly in [-alpha, alpha], but alpha is not reported (Section 2.1).
  • Partial update interval = 70 time steps (0.35 time units)
    Tested values 30, 70, 90; 70 used for main results (Section 6).
  • NRMSE threshold (epsilon) = 0.4
    Arbitrary tolerance for prediction horizon; no justification (Section 2.3).
assumptions (6)
  • domain assumption The focusing nonlinear Schrödinger equation (Eq. 11) is an accurate model for rogue wave dynamics in deep water and optics.
    Motivates the entire surrogate modeling exercise (Section 1).
  • domain assumption The Clean Numerical Simulation with Taylor expansion and quadruple precision yields ground truth that represents true NLS dynamics, and truncation to double precision does not alter the conclusions.
    Appendix A; the RC is trained and tested against this simulation.
  • standard math The reservoir is assumed to have the echo state property, so hidden state dependence on initial conditions washes out after Nw washout steps.
    Section 2.1, Eq. (3); relies on Jaeger's echo state property [22].
  • domain assumption A single shared output matrix Wo across all parallel reservoirs (translation invariance) is sufficient to capture the NLS dynamics on a uniform periodic grid.
    Section 2.2; justified by spatial invariance, following Barbosa et al. [45].
  • ad hoc to paper Rescaling the global L2 norm to its initial value at each step (Eq. 14) is a sufficient correction for the reservoir's dissipative energy drift.
    Section 6; this is an empirical fix, not derived from the Hamiltonian structure, and may not preserve other invariants.
  • domain assumption The local maximum stretching number computed with piecewise-constant coefficients is a valid estimate of the local Lyapunov exponent for scaling prediction horizons.
    Appendix B; used to interpret prediction horizon in Section 3.3.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Model-free Forecasting of Rogue Waves using Reservoir Computing." pith.science (2026). https://pith.science/paper/5GKPPMZC

@misc{pith2026250621918,
  author       = {Pith},
  title        = {Pith review of: Model-free Forecasting of Rogue Waves using Reservoir Computing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5GKPPMZC}},
  note         = {Machine review of arXiv:2506.21918}
}
read the original abstract

Recent research has demonstrated Reservoir Computing's capability to model various chaotic dynamical systems, yet its application to Hamiltonian systems remains relatively unexplored. This paper investigates the effectiveness of Reservoir Computing in capturing rogue wave dynamics from the nonlinear Schr\"{o}dinger equation, a challenging Hamiltonian system with modulation instability. The model-free approach learns from breather simulations with five unstable modes. A properly tuned parallel Echo State Network can predict dynamics from two distinct testing datasets. The first set is a continuation of the training data, whereas the second set involves a higher-order breather. An investigation of the one-step prediction capability shows remarkable agreement between the testing data and the models. Furthermore, we show that the trained reservoir can predict the propagation of rogue waves over a relatively long prediction horizon, despite facing unseen dynamics. Finally, we introduce a method to significantly improve the Reservoir Computing prediction in autonomous mode, enhancing its long-term forecasting ability. These results advance the application of Reservoir Computing to spatio-temporal Hamiltonian systems and highlight the critical importance of phase space coverage in the design of training data.

Figures

Figures reproduced from arXiv: 2506.21918 by the authors.

Figure 1
Figure 1. Illustration of the parallel RCs architecture. Th [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The NLS recurrence as training and testing data. Th [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Trained RC has low NRMSE for one-step forecasting. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: The prediction horizon (PH) of autonomous predict [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Autonomous RC_AB simulation for offset time instances. Actual data are plotted for comparison. Note how the amplitude difference diverges. Nevertheless, RC_AB can forecast the dynamics. The autonomous prediction starts at t0 = 13.4. The green curves are from RC_Cm, see…
Figure 6
Figure 6. Figure 6: Kinetic Energy (KE), Potential Energy (PE), and Ha [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 4
Figure 4. Figure 4: Overall, the prediction horizon in the maximum inten [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 7
Figure 7. Figure 7: A fifth-order maximum intensity breather as testin [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: RC_AB can accurately forecast the dynamics of maxi [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: RC can do a short time forecasting of propagation of [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: RC can capture the location of rogue waves in rando [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: Autonomous prediction for random ocean waves. Th [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: Heatmap of rogue wave dynamics. The upper left pan [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]
Figure 13
Figure 13. Figure 13: Sensitivity of the update interval: 30 (blue), 70 [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

70 extracted references · 62 canonical work pages

  1. [1]

    P . C. Liu, A chronology of freaque wave encounters, Geofiz ika 24 (1) (2007) 57 – 70. URL https://hrcak.srce.hr/file/2720

    work page 2007

  2. [2]

    Didenkulova, Catalogue of rogue waves occurred in the World Ocean from 2011 to 2018 reported by mass media sources, Ocean and Coastal Management 188 (2020) 105076

    E. Didenkulova, Catalogue of rogue waves occurred in the World Ocean from 2011 to 2018 reported by mass media sources, Ocean and Coastal Management 188 (2020) 105076. doi:10.1016 /j.ocecoaman.2019.105076

    arXiv 2020

  3. [3]

    Kharif, E

    C. Kharif, E. Pelinovsky, Physical mechanisms of the rog ue wave phenomenon, European Journal of Mechanics - B /Fluids 22 (6) (2003) 603–634. doi:10.1016 /j.euromechflu.2003.09.002

    work page 2003

  4. [4]

    Christou, K

    M. Christou, K. Ewans, Field Measurements of Rogue Water Waves, Journal of Physical Oceanography 44 (9) (2014) 2317– 2335. doi:10.1175/JPO-D-13-0199.1

    work page doi:10.1175/jpo-d-13-0199.1 2014

  5. [5]

    Fedele, J

    F. Fedele, J. Brennan, S. Ponce de Leon, J. Dudley, F. Dias , Real world ocean rogue waves explained without the modulat ional instability, Scientific Reports 6 (1) (2016) 27715. doi:10.1038 /srep27715

    work page 2016

  6. [6]

    Onorato, A

    M. Onorato, A. R. Osborne, M. Serio, S. Bertone, Freak Wav es in Random Oceanic Sea States, Physical Review Letters 86 ( 25) (2001) 5831–5834. doi:10.1103 /PhysRevLett.86.5831

    work page 2001

  7. [7]

    Onorato, A

    M. Onorato, A. R. Osborne, M. Serio, Modulational Instab ility in Crossing Sea States: A Possible Mechanism for the Fo rmation of Freak Waves, Physical Review Letters 96 (1) (2006) 014503. doi:10 .1103/PhysRevLett.96.014503

    work page 2006

  8. [8]

    Onorato, D

    M. Onorato, D. Proment, A. To ffoli, Freak waves in crossing seas, The European Physical Jou rnal Special Topics 185 (1) (2010) 45–55. doi:10.1140/epjst/e2010-01237-8

    work page doi:10.1140/epjst/e2010-01237-8 2010

Show all 70 references
  1. [9]

    Akhmediev, J

    N. Akhmediev, J. M. Soto-Crespo, A. Ankiewicz, Extreme w aves that appear from nowhere: On the nature of rogue waves, P hysics Letters, Section A: General, Atomic and Solid State Physics 373 (25) ( 2009) 2137–2145. doi:10.1016 /j.physleta.2009.04.023

    2009

  2. [10]

    W. M. Moslem, P . K. Shukla, B. Eliasson, Surface plasma r ogue waves, Europhysics Letters 96 (2) (2011) 25002. doi:10 .1209/0295- 5075/96/25002

    2011

  3. [11]

    G. P . V eldes, J. Borhanian, M. McKerr, V . Saxena, D. J. Frantzeskakis, I. Kourakis, Electromagnetic rogue waves in beam–plasma interactions, Journal of Optics 15 (6) (2013) 064003. doi:10.1088 /2040-8978/15/6/064003

    2013

  4. [12]

    Y ang, Y .-X

    Y . Y ang, Y .-X. Gao, H.-W. Y ang, Analysis of the rogue waves in the blood based on the high-order NLS equations with var iable coefficients, Chinese Physics B 30 (11) (2021) 110202. doi:10.1088 /1674-1056/abff31

    2021

  5. [13]

    Li, Z.-J

    J. Li, Z.-J. Y ang, S.-M. Zhang, Characteristics of cert ain higher-order Hermite-cos-Gauss breathing solitons in duced by the initial wavefront bending in optical media with nonlocal nonlinearity, Chaos , Solitons & Fractals 187 (2024) 115338. doi:10.1016 /j.chaos.2024.115338

    2024

  6. [14]

    Y . Liu, Z. Zhao, Rogue waves of the (2 + 1)-dimensional integrable reverse space–time nonlocal Sc hrödinger equation, Theoretical and Mathematical Physics 222 (1) (2025) 34–52. doi:10.1134 /S0040577925010040

    2025

  7. [15]

    A. V . Slunyaev, D. E. Pelinovsky, E. N. Pelinovsky, Rogue waves in the sea: Observations, physics, and mathematics, Physics-Uspekhi 66 (2) (2023) 148–172. doi:10.3367 /UFNe.2021.08.039038. 23 Hasmi & Susanto / Communications in Nonlinear Science and Numerical Simulat ion 00 (...

    2023

  8. [16]

    Klein, M

    M. Klein, M. Dudek, G. F. Clauss, S. Ehlers, J. Behrendt, N. Hoffmann, M. Onorato, On the Deterministic Prediction of Water W aves, Fluids 5 (1) (2020) 9. doi:10.3390 /fluids5010009

    2020

  9. [17]

    Cousins, M

    W. Cousins, M. Onorato, A. Chabchoub, T. P . Sapsis, Pred icting ocean rogue waves from point measurements: An experi mental study for unidirectional waves, Physical Review E 99 (2019) 032201. d oi:10.1103/PhysRevE.99.032201

    2019 doi

  10. [18]

    Farazmand, T

    M. Farazmand, T. P . Sapsis, Reduced-order prediction o f rogue waves in two-dimensional deep-water waves, Journal of Computational Physics 340 (2017) 418–434. doi:10.1016 /j.jcp.2017.03.054

    2017

  11. [19]

    Kagemoto, Forecasting a water-surface wave train wi th artificial intelligence- A case study, Ocean Engineering 207 (2020) 107380

    H. Kagemoto, Forecasting a water-surface wave train wi th artificial intelligence- A case study, Ocean Engineering 207 (2020) 107380. doi:10.1016/j.oceaneng.2020.107380

    2020

  12. [20]

    H. Kagemoto, Forecasting a water-surface wave train wi th artificial intelligence (Part 2) – Can the occurrence of fr eak waves be predicted with AI?, Ocean Engineering 252 (2022) 111205. doi:10.1016 /j.oceaneng.2022.111205

    2022

  13. [21]

    Breunung, B

    T. Breunung, B. Balachandran, Data-driven, high resol ution ocean wave forecasting and extreme wave predictions, Ocean Engineering 268 (2023) 113271. doi:10.1016 /j.oceaneng.2022.113271

    2023

  14. [22]

    echo state

    H. Jaeger, The "echo state" approach to analysing and tr aining recurrent neural networks – with an Erratum note, Tec h. rep., German National Research Center for Information Technology, (2001)

    2001

  15. [23]

    Jaeger, H

    H. Jaeger, H. Haas, Harnessing Nonlinearity: Predicti ng Chaotic Systems and Saving Energy in Wireless Communication, Science 304 (5667) (2004) 78–80. doi:10.1126 /science.1091277

    2004

  16. [24]

    Lukoševi ˇcius, H

    M. Lukoševi ˇcius, H. Jaeger, Reservoir computing approaches to recurre nt neural network training, Computer Science Review 3 (3) (2 009) 127–149. doi:10.1016 /j.cosrev.2009.03.005

    2009

  17. [25]

    Chattopadhyay, P

    A. Chattopadhyay, P . Hassanzadeh, D. Subramanian, Data-driven predictions of a multiscale Lorenz 96 chaotic system using machine-learning methods: Reservoir computing, artificial neural network, a nd long short-term memory network, Nonlinear Processes in G eophysics 27 (3) (20...

    2020

  18. [26]

    Pathak, Z

    J. Pathak, Z. Lu, B. R. Hunt, M. Girvan, E. Ott, Using mach ine learning to replicate chaotic attractors and calculate Lyapunov exponents from data, Chaos: An Interdisciplinary Journal of Nonlinea r Science 27 (12) (2017) 121102, publisher: American Instit ute of Physics. doi:...

    2017 doi

  19. [27]

    Pathak, B

    J. Pathak, B. Hunt, M. Girvan, Z. Lu, E. Ott, Model-Free P rediction of Large Spatiotemporally Chaotic Systems from D ata: A Reservoir Computing Approach, Physical Review Letters 120 (2) (2018) 24102. doi:10.1103 /PhysRevLett.120.024102

    2018

  20. [28]

    N. A. Doan, W. Polifke, L. Magri, Short- And long-term pr edictions of chaotic flows and extreme events: A physics-con strained reser- voir computing approach, Proceedings of the Royal Society A : Mathematical, Physical and Engineering Sciences 477 (202 1) 2253. doi:10.1098/rsp...

  21. [29]

    A. Röhm, D. J. Gauthier, I. Fischer, Model-free inferen ce of unseen attractors: Reconstructing phase space featur es from a single noisy trajectory using reservoir computing, Chaos 31 (2021) 1031 27. doi:10.1063 /5.0065813

    2021

  22. [30]

    D. J. Gauthier, I. Fischer, A. Röhm, Learning unseen coe xisting attractors, Chaos: An Interdisciplinary Journal o f Nonlinear Science 32 (11) (2022) 113107, publisher: American Institute of Physics. d oi:10.1063/5.0116784

    2022 doi

  23. [31]

    F. Huhn, L. Magri, Gradient-free optimization of chaot ic acoustics with reservoir computing, Physical Review Flu ids 7 (2022) 014402. doi:10.1103/PhysRevFluids.7.014402

    2022 doi

  24. [32]

    Jiang, Y

    J. Jiang, Y . C. Lai, Model-free prediction of spatiotem poral dynamical systems with recurrent neural networks: Ro le of network spectral radius, Physical Review Research 1 (2019) 033056. doi:10.1 103/PhysRevResearch.1.033056

    2019

  25. [33]

    Pandey, J

    S. Pandey, J. Schumacher, Reservoir computing model of two-dimensional turbulent convection, Physical Review Fl uids 5 (11) (2020) 113506. doi:10.1103 /PhysRevFluids.5.113506

    2020

  26. [34]

    K. Y ao, E. Forgoston, P . Y ecko, Learning ocean circulat ion models with reservoir computing, Physics of Fluids 34 (1 1) (2022) 116604. doi:10.1063/5.0119061

    2022 doi

  27. [35]

    Bonas, C

    M. Bonas, C. K. Wikle, S. Castruccio, Calibrated foreca sts of quasi-periodic climate processes with deep echo stat e networks and penalized quantile regression, Environmetrics 35 (1) (2024) e2833. d oi:10.1002/env.2833. URL https://onlinelibrary.wiley.com/doi/abs/10.1002/env.2833

    2024 doi

  28. [36]

    Pathak, A

    J. Pathak, A. Wikner, R. Fussell, S. Chandra, B. R. Hunt, M. Girvan, E. Ott, Hybrid forecasting of chaotic processes: Using machine learning in conjunction with a knowledge-based model, Chaos: An Inte rdisciplinary Journal of Nonlinear Science 28 (4) (2018) 04 1101, publisher:...

    2018

  29. [37]

    Wikner, J

    A. Wikner, J. Pathak, B. Hunt, M. Girvan, T. Arcomano, I. Szunyogh, A. Pomerance, E. Ott, Combining machine learning with knowledge- based modeling for scalable forecasting and subgrid-scale closure of large, complex, spatiotemporal systems, Chaos 3 0 (5) (2020) 053111. doi:10...

    2020 doi

  30. [38]

    Tanaka, T

    G. Tanaka, T. Y amane, J. B. Héroux, R. Nakane, N. Kanazaw a, S. Takeda, H. Numata, D. Nakano, A. Hirose, Recent advance s in physical reservoir computing: A review, Neural Networks 115 (2019) 1 00–123. arXiv:1808.04962, doi:10.1016 /j.neunet.2019.03.005

    2019 arXiv

  31. [39]

    Liang, J

    X. Liang, J. Tang, Y . Zhong, B. Gao, H. Qian, H. Wu, Physic al reservoir computing with emerging electronics, Nature E lectronics 7 (3) (2024) 193–206. doi:10.1038 /s41928-024-01133-z

    2024

  32. [40]

    Everschor-Sitte, A

    K. Everschor-Sitte, A. Majumdar, K. Wolk, D. Meier, Top ological magnetic and ferroelectric systems for reservoir computing, Nature Re- views Physics 6 (7) (2024) 455–462. doi:10.1038 /s42254-024-00729-w

    2024

  33. [41]

    Zhang, H

    H. Zhang, H. Fan, L. Wang, X. Wang, Learning Hamiltonian dynamics with reservoir computing, Physical Review E 104 (2 ) (2021) 024205. doi:10.1103/PhysRevE.104.024205

    2021 doi

  34. [42]

    Pershin, C

    A. Pershin, C. Beaume, K. Li, S. M. Tobias, Training a neu ral network to predict dynamics it has never seen, Physical R eview E 107 (1) (2023) 014304. doi:10.1103 /PhysRevE.107.014304

    2023

  35. [43]

    J. Z. Kim, Z. Lu, E. Nozari, G. J. Pappas, D. S. Bassett, Te aching recurrent neural networks to infer global temporal s tructure from local examples, Nature Machine Intelligence 3 (4) (2021) 316–323 . doi:10.1038 /s42256-021-00321-2

    2021

  36. [44]

    Manjunath, H

    G. Manjunath, H. Jaeger, Echo State Property Linked to a n Input: Exploring a Fundamental Characteristic of Recurre nt Neural Networks, Neural Computation 25 (3) (2013) 671–696. doi:10.1162 /NECO_a_0411

    2013

  37. [45]

    W. A. Barbosa, A. Gri ffith, G. E. Rowlands, L. C. Govia, G. J. Ribeill, M. H. Nguyen, T. A. Ohki, D. J. Gauthier, Symmetry-aware reservoir 24 Hasmi & Susanto / Communications in Nonlinear Science and Numerical Simulat ion 00 (2025) 1–25 25 computing, Physical Review E 104 (2021...

    2025

  38. [46]

    R. Pyle, N. Jovanovic, D. Subramanian, K. V . Palem, A. B. Patel, Domain-driven models yield better predictions at lo wer cost than reser- voir computers in Lorenz systems, Philosophical Transacti ons of the Royal Society A: Mathematical, Physical and Engin eering Sciences 379...

    2021

  39. [47]

    N. N. Akhmediev, V . M. Eleonskii, N. E. Kulagin, Exact fir st-order solutions of the nonlinear Schrödinger equation, Theoretical and Mathe- matical Physics 72 (2) (1987) 809–818. doi:10.1007 /BF01017105

    1987

  40. [48]

    B. M. Herbst, M. J. Ablowitz, Numerically induced chaos in the nonlinear schrödinger equation, Physical Review Let ters 62 (18) (1989) 2065–2068. doi:10.1103 /PhysRevLett.62.2065

    1989

  41. [49]

    M. R. Beli ´c, S. N. Nikoli ´c, O. A. Ashour, N. B. Aleksi ´c, On di fferent aspects of the optical rogue waves nature, Nonlinear D ynamics 108 (2) (2022) 1655–1670. doi:10.1007 /s11071-022-07284-y

    2022

  42. [50]

    T. Hu, S. Liao, On the risks of using double precision in n umerical simulations of spatio-temporal chaos, Journal of Computational Physics 418 (2020) 109629. doi:10.1016 /j.jcp.2020.109629

    2020

  43. [51]

    H. C. Y uen, W. E. Ferguson, Relationship between Benjam in-Feir instability and recurrence in the nonlinear Schröd inger equation, Physics of Fluids 21 (8) (1978) 1275–1278. doi:10.1063 /1.862394

    1978

  44. [52]

    Erkintalo, K

    M. Erkintalo, K. Hammani, B. Kibler, C. Finot, N. Akhmed iev, J. M. Dudley, G. Genty, Higher-order modulation instability in nonlinear fiber optics, Physical Review Letters 107 (25) (2011) 14–18. doi: 10.1103/PhysRevLett.107.253901

    2011 doi

  45. [53]

    S. A. Chin, O. A. Ashour, M. R. Beli ´c, Anatomy of the Akhmediev breather: Cascading instability, first formation time, and Fermi-Pasta-Ulam recurrence, Physical Review E 92 (6) (2015) 063202. doi:10. 1103/PhysRevE.92.063202

    2015

  46. [54]

    M. Roy, S. Mandal, C. Hens, A. Prasad, N. V . Kuznetsov, M. Dev Shrimali, Model-free prediction of multistability usi ng echo state network, Chaos: An Interdisciplinary Journal of Nonlinear Science 3 2 (10) (2022) 101104. doi:10.1063 /5.0119963

    2022

  47. [55]

    S. A. Chin, O. A. Ashour, S. N. Nikoli ´c, M. R. Beli ´c, Maximal intensity higher-order Akhmediev breathers of t he nonlinear Schrödinger equation and their systematic generation, Physics Letters , Section A: General, Atomic and Solid State Physics 380 (43) (2016) 3625–3629. ...

    2016 doi

  48. [56]

    Akhmediev, N

    N. Akhmediev, N. Mitzkevich, Extremely high degree of N -soliton pulse compression in an optical fiber, IEEE Journal of Quantum Electronics 27 (3) (1991) 849–857. doi:10.1109 /3.81399

    1991

  49. [57]

    D. J. Kedziora, A. Ankiewicz, N. Akhmediev, Circular ro gue wave clusters, Physical Review E - Statistical, Nonline ar, and Soft Matter Physics 84 (5) (2011) 056611. doi:10.1103 /PhysRevE.84.056611

    2011

  50. [58]

    H. Fan, J. Jiang, C. Zhang, X. Wang, Y . C. Lai, Long-term p rediction of chaotic systems with machine learning, Physic al Review Research 2 (1) (2020) 012080. doi:10.1103 /PhysRevResearch.2.012080

    2020

  51. [59]

    T. P . Sapsis, Statistics of Extreme Events in Fluid Flow s and Waves, Annual Review of Fluid Mechanics 53 (2021) 85–11 1. doi:10.1146/annurev-fluid-030420-032810

    2021 doi

  52. [60]

    Fujii, K

    K. Fujii, K. Nakajima, Harnessing Disordered-Ensembl e Quantum Dynamics for Machine Learning, Physical Review Ap plied 8 (2) (2017) 024030. doi:10.1103 /PhysRevApplied.8.024030

    2017

  53. [61]

    W. A. S. Barbosa, D. J. Gauthier, Learning spatiotempor al chaos using next-generation reservoir computing, Chaos : An Interdisciplinary Journal of Nonlinear Science 32 (9) (2022) 093137, publishe r: American Institute of Physics. doi:10.1063 /5.0098707

    2022

  54. [62]

    P . J. Ehlers, H. I. Nurdin, D. Soh, Stochastic reservoir computers, Nature Communications 16 (1) (2025) 3070. doi:1 0.1038/s41467-025- 58349-6

    2025

  55. [63]

    Köster, K

    F. Köster, K. Kanno, A. Uchida, Attention-Enhanced Res ervoir Computing as a Multiple Dynamical System Approximat or (2025). arXiv:2505.05852, doi:10.48550/arXiv.2505.05852

    2025 doi arXiv

  56. [64]

    V oglis, G

    N. V oglis, G. J. Contopoulos, Invariant spectra of orbi ts in dynamical systems, Journal of Physics A: Mathematical and General 27 (14) (1994) 4899. doi:10.1088 /0305-4470/27/14/017

    1994

  57. [65]

    Eckhardt, D

    B. Eckhardt, D. Y ao, Local Lyapunov exponents in chaoti c systems, Physica D: Nonlinear Phenomena 65 (1) (1993) 100– 108. doi:10.1016/0167-2789(93)90007-N

    1993 doi

  58. [66]

    Skokos, E

    C. Skokos, E. Gerlach, Numerical integration of variat ional equations, Physical Review E 82 (3) (2010) 036704. doi:10.1103/PhysRevE.82.036704

    2010 doi

  59. [67]

    Skokos, The Lyapunov Characteristic Exponents and T heir Computation, in: Dynamics of Small Solar System Bodies and Exoplanets, Springer, Berlin, Heidelberg, 2010, pp

    C. Skokos, The Lyapunov Characteristic Exponents and T heir Computation, in: Dynamics of Small Solar System Bodies and Exoplanets, Springer, Berlin, Heidelberg, 2010, pp. 63–135. doi:10.10 07/978-3-642-04458-8_2

    2010

  60. [68]

    Kharif, E

    C. Kharif, E. Pelinovsky, A. Slunyaev, Rogue Waves in th e Ocean, Advances in Geophysical and Environmental Mechanics and Mathematics, Springer, Berlin, Heidelberg, 2009. doi:10.1007 /978-3-540-88419-4

    2009

  61. [69]

    Hasselmann, T

    K. Hasselmann, T. Barnett, E. Bouws, H. Carlson, D. Cart wright, K. Enke, J. Ewing, H. Gienapp, D. Hasselmann, P . Kruseman, A. Meerburg, P . Muller, D. Olbers, K. Richter, W. Sell, H. Walden, Measure ments of wind-wave growth and swell decay during the joint no rth sea wave p...

    1973

  62. [70]

    Mackay, A unified model for unimodal and bimodal ocean wave spectra, International Journal of Marine Energy 15 (20 16) 17–40

    E. Mackay, A unified model for unimodal and bimodal ocean wave spectra, International Journal of Marine Energy 15 (20 16) 17–40. doi:10.1016/j.ijome.2016.04.015. 25

    2016 doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.