Pith. sign in

REVIEW 3 major objections 5 minor 61 references

In search of rogue waves: a novel proposal distribution for parallelized rejection sampling of the truncated KdV Gibbs measure

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A paper introduces a proposal distribution that makes rejection sampling of the truncated KdV Gibbs measure up to six orders of magnitude more efficient and nearly perfectly parallel.

desk verdict A useful rejection-sampling trick for the TKdV Gibbs measure with credible large speedups, but the 'exact' claim rests on an unverified global maximum. read the letter →

arxiv 2411.16952 v2 pith:Z3GIICGS submitted 2024-11-25 math.NA cs.NAphysics.data-an

Nicholas J. Moore , Brendan Foerster This is my paper · ORCID
classification math.NAcs.NAphysics.data-an MSC 65C0565Y0576B1582B31
keywords truncatedKdVGibbsmeasurerejectionsamplingproposaldistributionroguewavesanisotropicGaussianparallelextremewavestatistics
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 aims to make rejection sampling of the truncated KdV Gibbs measure practical in the parameter regime where anomalous, extreme waves occur. It replaces the spectrally uniform proposal with an anisotropic Gaussian distribution derived from the linearized system, which is nearly as easy to sample but much closer to the target measure. The paper reports that this proposal yields 1 to 6 orders of magnitude more accepted samples than the naive uniform proposal, with acceptance rates above 80% in the linear case and improvements as large as 2.5 million-fold in the strongest tested case. If correct, the sampler gives independent, uncorrelated wave-field samples at high throughput and with near-optimal parallel speedup, making it a practical tool for studying rogue-wave statistics.

What carries the argument

The machinery is the proposal density $g$ on the unit sphere $S^{2K-1}$, obtained by projecting an anisotropic Gaussian with variances $\sigma_k^2 = 1/(1+\alpha^*\beta' k^2/K^3)$, where the self-consistent parameter $\alpha^*$ is the unique root of $F(\alpha) = 1 - (\alpha/K) \sum_{k=1}^K 1/(1+\alpha\beta' k^2/K^3)$. Drawing from $g$ is one Gaussian draw plus a radial projection, and the acceptance step uses the closed-form ratio $f(\hat x)/g(\hat x) = \exp(-\beta H_K[\hat U(\hat x)]) \left(1 + (\alpha^* \beta'/K^3) \sum_k k^2(\hat x_k^2 + \hat x_{K+k}^2)\right)^K$. $H_2$ and $H_3$ are evaluated in $O(K^2)$ time, with $H_3$ computed by a double-summation rearrangement of the triple-sum formula, and the rejection constant $M$ is chosen by Nelder-Mead maximization of the ratio, initialized at the Dirichlet kernel.

What would settle it

For the parameters in Tables I and II, scan the sphere densely (millions of independent draws from the proposal plus a multi-start global optimizer) and check whether any state yields $f(\hat x)/g(\hat x)$ greater than the reported $M$; if such a state exists, the acceptance probability would exceed 1 there and the sampler is biased. A cheaper check is to compare ensemble statistics (mean power spectrum, skewness) against a long independent MCMC run: disagreement beyond Monte Carlo error would indicate a misspecified rejection constant.

Watch

Extended reading notes

Core claim

The central claim is that the Gibbs density $f$ of the TKdV system, supported on the energy sphere, can be sampled exactly by rejection when the proposal is the density $g$ induced by an anisotropic Gaussian with per-mode variances $\sigma_k^2 = 1/(1+\alpha^* \beta' k^2/K^3)$. Because $g$ was constructed to match the linearized ($C_3=0$) Gibbs measure in the large-$K$ limit, the ratio $f/g$ stays close to constant, and the acceptance probability $f/(M g)$ is high; numerical tests show acceptance rates above 80% for linear TKdV and improvements over the naive uniform proposal of 1--6 orders of magnitude in the skewness-producing regimes where extreme waves appear. The sampler also reproduces the expected physics: linear cases give symmetric, nearly Gaussian surface displacements, while increasing the nonlinearity ratio $C_3/C_2$ produces positive skewness, and 1-in-500 sampled wave fields exceed the $4\sigma$ rogue-wave threshold by up to 24%.

Load-bearing premise

The load-bearing premise is that the Nelder-Mead search starting from the Dirichlet kernel finds the true global maximum of the ratio $f/g$ on the sphere, since no proof or verification of global optimality is provided for the rejection constant $M$.

Editorial extensions

If this is right

  • For linear TKdV the acceptance rate exceeds 80%, so the sampler is nearly free in that regime.
  • In the nonlinear regimes where 1-in-500 events exceed the $4\sigma$ threshold, the improved proposal still beats the uniform one by 1--3 orders of magnitude, making extreme-event searches feasible.
  • Because accepted samples are independent, the algorithm parallelizes with essentially no communication, achieving 80% of optimal speedup on 16 cores.
  • The method yields exact samples from the Gibbs measure, so histograms, spectra, and extremal statistics computed from them are unbiased.
  • The same proposal can be used in sampling-importance resampling, with no need for the global rejection constant.

Reading between the lines

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

  • A natural extension is to add a cubic tilt to the proposal to match the skewness contribution of $H_3$, which would raise acceptance rates in the high-nonlinearity regime where the current method degrades.
  • The non-monotone improvement factors seen at $K=32$ suggest the ratio $f/g$ has exploitable structure, so a tighter analytic bound on $M$ may be derivable.
  • The sampler's independence makes it a clean generator of training data for machine-learning models that flag pre-rogue states, a direction the paper lists as future work.
  • Because $g$ is derived from the linear system but works beyond it, the approach may transfer to other Hamiltonian truncations where a quadratic energy dominates the Gibbs weight.
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

3 major / 5 minor

Summary. The paper proposes a rejection sampling algorithm for the mixed canonical-microcanonical Gibbs measure of the truncated KdV equation. The proposal is an anisotropic Gaussian distribution projected onto the energy sphere, originally constructed by Sun and Moore (2023) to approximate the linear TKdV measure. The authors derive a closed-form expression for the proposal density g (Eq. 38), define a rejection sampler with constant M (Eq. 45), and report acceptance rates and improvements over uniform proposals for K=16 and K=32, parallel speedup, and examples of extreme wave fields. The central claim is that within experimentally relevant parameter regimes the proposal yields 1--6 orders of magnitude more accepted samples than the uniform proposal.

Significance. If the algorithm is correct as stated, it is a valuable contribution: it produces independent samples from the target measure, is trivially parallelizable, and avoids MCMC tuning and burn-in. The use of a self-consistently chosen alpha* (Eq. 29) and the closed-form proposal density are strong points, and the acceptance-rate comparison to a uniform proposal is a fresh numerical contribution. The paper is honest about the heuristic nature of the optimization step, but that heuristic is load-bearing because the exactness of rejection sampling depends on a valid upper bound M.

major comments (3)
  1. [IV.A, Eqs. (44)-(45) and (47); Appendix A] The rejection constant M is found by Nelder-Mead from a Dirichlet-kernel initial guess, with no global optimality certificate. This is load-bearing: if M is underestimated, the acceptance probability f/(Mg) exceeds 1 for some states; accepting those with probability 1 biases the sampler and inflates the reported acceptance rates. The concern is not purely formal: for K=2, Appendix A gives H3 = 2π R2 R1^2 cos(2θ1−θ2), whose maximizer under the energy constraint has R2/R1 = 1/√2, while Eq. (47) has equal amplitudes R2/R1 = 1. Thus the initial guess is not the maximizer of the dominant factor in Eq. (46), even in the simplest case, and a single Nelder-Mead run in dimensions up to 63 does not establish the required upper bound. Please add a numerical certificate (e.g., multi-start or global optimization, or verification that max f/(Mg) ≤ 1 over a very large proposal sample) or a rigorous upper bound for Eq. (40).
  2. [V.B, Tables I and II] The acceptance-rate measurements lack error bars, and the extreme entries are based on very few accepted samples. For example, Table II reports acceptance rates as low as 2.1×10^{-9}, and the text states that for the asterisked cases the naive version could not produce 100 samples over weeks; the corresponding improvement factors are thus point estimates with unknown relative error. Since the central quantitative claim is the improvement factor (1–6 orders of magnitude), please report confidence intervals or bootstrap estimates, or at least the number of accepted samples underlying each entry. This is necessary for the reader to judge whether the 'orders of magnitude' statement holds in the extreme nonlinear regime.
  3. [V.B, Algorithm 1 and Section IV.B] The paper asserts that the algorithm generates independent samples from the target measure, but the only evidence is the internal acceptance-rate computation; there is no independent check of the output distribution. A direct comparison of sampled moments (e.g., skewness and power spectra) against a long MCMC or SIR run in a few moderate cases would provide a strong safeguard against undetected bias caused by an invalid M. Without such a check, the exactness claim rests entirely on the unverified upper bound in Eq. (45).
minor comments (5)
  1. [Abstract and Introduction] There is a grammatical error: 'include better parallelization properties' should be 'including better parallelization properties.'
  2. [III.B, Eq. (36)] In the Laplace transform formula L[t^{K-1}] = ∫ exp(-st) t^{K-1} dt, the integration variable is t (or u after the change of variables), not du; the displayed notation is inconsistent and should be corrected.
  3. [II.C and III.A] The notation Const is used repeatedly for different constants in Eqs. (27), (34), (37), (38), and (46). This is acceptable informally, but the authors should add a sentence reminding the reader that these constants are not equal, to avoid confusion in the ratio f/g derivation.
  4. [V.C] The parallel speedup is demonstrated on a single hardware platform (Apple M2 Ultra) with no details on how the timing was measured or whether the speedup curve is averaged over repeated runs. A brief description of the measurement procedure and variance would strengthen this section.
  5. [Reproducibility] No code or data are provided. Since the rejection constant M is computed by a numerical optimization and the reported acceptance rates are stochastic estimates, supplying code or pseudorandom seeds would greatly improve reproducibility.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the proposal distribution is imported from a same-author prior paper, but the acceptance-rate improvements are fresh numerical measurements and the sampler's exactness does not reduce to that citation.

full rationale

The derivation chain is not circular. The proposal density g is defined directly by Eqs. (25)-(27) and (38), and the target density f is defined by the Gibbs measure in Eq. (19); the ratio f/g is computed exactly in Eq. (40). The rejection constant M is defined as a maximum in Eq. (45) and estimated numerically; this is a global-optimization risk, not a self-referential reduction. The convergence statement Eq. (30) is imported from Sun & Moore (2023), a paper co-authored by N. J. Moore, and it motivates the proposal, but the reported 1-6 orders-of-magnitude improvement is measured directly in Tables I-II rather than derived from that theorem. Nor is alpha* fitted to the reported acceptance rates: Eq. (29) is a self-consistency condition solved independently of the performance tests. No fitted parameter is renamed as a prediction, and no uniqueness theorem is invoked to force the choice of g. The claimed advantage over uniform rejection sampling is therefore an empirical, falsifiable measurement, not a conclusion that is equivalent to its input. The only notable caveat, the absence of a global-maximum certificate for M, is a correctness risk rather than a circularity.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The central sampler rests on the TKdV Gibbs measure from prior work, on the Sun-Moore convergence theorem, also from the same group, and on the heuristic numerical maximization of the ratio f/g. alpha* is an auxiliary parameter fixed by a scalar root equation, not fitted to data. No new physical entities are introduced.

free parameters (1)
  • alpha* = root of Eq. (29), about 1.2 for K=32, beta'=20
    Introduced as an extra degree of freedom in the change of measure; selected self-consistently as the root of F(alpha)=0, not fitted to data. It controls the covariance scale of the proposal distribution and enters g via Eq. (38).
assumptions (4)
  • domain assumption The mixed canonical-microcanonical Gibbs measure dG proportional to exp(-beta H) delta(E-E0) is the correct target distribution for TKdV extreme wave statistics.
    Adopted from prior TKdV statistical mechanics [11,16,28], not re-derived; the paper samples this measure rather than proving it.
  • domain assumption The theorem of Sun and Moore (2023) that dgammaSigma converges to the linear TKdV Gibbs measure as K tends to infinity (Eq. (30)) is valid.
    Used to motivate g as approximating f; the proof is cited but not reproduced.
  • ad hoc to paper The Dirichlet kernel (Eq. (47)) provides a sufficient initial guess for the global maximization of f/g, and Nelder-Mead converges to the global maximum M.
    No proof that the optimization finds the true supremum in Eq. (45); if M is underestimated, acceptance probabilities can exceed 1.
  • standard math H3 can be computed exactly via Eq. (42), and the truncation projection is handled correctly.
    Standard algebra of Fourier modes; validated in Appendix A for K=2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of In search of rogue waves: a novel proposal distribution for parallelized rejection sampling of the truncated KdV Gibbs measure." pith.science (2026). https://pith.science/paper/Z3GIICGS

@misc{pith2026241116952,
  author       = {Pith},
  title        = {Pith review of: In search of rogue waves: a novel proposal distribution for parallelized rejection sampling of the truncated KdV Gibbs measure},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z3GIICGS}},
  note         = {Machine review of arXiv:2411.16952}
}
read the original abstract

The Gibbs ensemble of the truncated KdV (TKdV) equation has been shown to accurately describe the anomalous wave statistics observed in laboratory experiments, in particular the emergence of extreme events. Here, we introduce a novel proposal distribution that facilitates efficient rejection sampling of the TKdV Gibbs measure. Within parameter regimes accessible to laboratory experiments and capable of producing extreme events, the proposal distribution generates 1-6 orders of magnitude more accepted samples than does a naive, uniform distribution. When equipped with the new proposal distribution, a simple rejection algorithm enjoys key advantages over a Markov chain Monte Carlo algorithm, include better parallelization properties and generation of uncorrelated samples.

Figures

Figures reproduced from arXiv: 2411.16952 by the authors.

Figure 1
Figure 1. provides a visualization of how α ∗ is selected and how its value depends on the cutoff wavenumber K [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2: Visualization of the Dirichlet kernel for [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3: Visualizing wavefield statistics in the case of linear TKdV ( [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: FIG. 4: The effects of strong nonlinearity, [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5: The parallel speedup of the rejection algorithm versus the number of processors. Test conducted [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6: Sampling individual wave fields to search for extreme events. For each [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7: Same test as in Fig. 6 except with [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8: Sampling of individual wave fields with [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9: Same test as in Fig. 8 except with larger inverse temperature [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 61 canonical work pages

  1. [1]

    (45) using Eq

    Precompute the the rejection constant M by numerically maximizing Eq. (45) using Eq. (47) as the initial guess

    work page

  2. [2]

    (25) and (26) with α = α∗ selected as the root of Eq

    Draw a sample X ∈ R2K from the anisotropic Gaussian distribution dγΣ given by Eqs. (25) and (26) with α = α∗ selected as the root of Eq. (29)

    work page

  3. [3]

    By doing so, ˆX represents a random sample drawn from density g

    Project the sample X onto S2K−1 via ˆX = X/ |X|. By doing so, ˆX represents a random sample drawn from density g

    work page

  4. [4]

    (40), with Eqs

    For the sample ˆX, compute the acceptance probability f ( ˆX)/(M g( ˆX)) using Eq. (40), with Eqs. (20) and (42) used to compute H2 and H3. Decide to accept the sample or not based on the computed probability. If accepted, ˆX represents a random sample drawn from density f . Through Eq. (16), an accepted ˆX produces a Fourier-coefficient set, ˆuk for k = ...

    work page

  5. [5]

    We note that it is trivial to parallelize the above algorithm

    Go to step 2 and repeat until a desired number of samples have been accepted. We note that it is trivial to parallelize the above algorithm. In particular, step 1 should be precomputed and then steps 2–5 can be performed in parallel with no communication required between computational nodes. In the following section, we report numerical results on both se...

    work page 2023

  6. [6]

    Dematteis, T

    G. Dematteis, T. Grafke, M. Onorato, and E. Vanden-Eijnden, Experimental evidence of hydrodynamic instan- tons: the universal route to rogue waves, Physical Review X 9, 041057 (2019)

    work page 2019

  7. [7]

    Haver, A possible freak wave event measured at the Draupner Jacket January 1 1995, in Rogue waves, Vol

    S. Haver, A possible freak wave event measured at the Draupner Jacket January 1 1995, in Rogue waves, Vol. 2004 (2004) pp. 1–8

    work page 2004

  8. [8]

    M¨ uller, C

    P. M¨ uller, C. Garrett, and A. Osborne, Rogue waves: The fourteenth ’aha ’huliko’a hawaiian winter workshop, Oceanography 18 (2005)

    work page 2005

Show all 61 references
  1. [9]

    Garrett and J

    C. Garrett and J. Gemmrich, Rogue waves, Phys. Today 62, 62 (2009)

    2009

  2. [10]

    Hadjihosseini, J

    A. Hadjihosseini, J. Peinke, and N. Hoffmann, Stochastic analysis of ocean wave states with and without rogue waves, New Journal of Physics 16, 053037 (2014)

    2014

  3. [11]

    Dematteis, T

    G. Dematteis, T. Grafke, and E. Vanden-Eijnden, Rogue waves and large deviations in deep sea, Proceedings of the National Academy of Sciences 115, 855 (2018)

    2018

  4. [12]

    A. J. Majda and D. Qi, Statistical phase transitions and extreme events in shallow water waves with an abrupt depth change, Journal of Statistical Physics 179, 1718 (2020)

    2020

  5. [13]

    C. T. Bolles, K. Speer, and M. Moore, Anomalous wave statistics induced by abrupt depth change, Physical Review Fluids 4, 011801 (2019)

    2019

  6. [14]

    J. M. Dudley, G. Genty, A. Mussot, A. Chabchoub, and F. Dias, Rogue waves and analogies in optics and oceanography, Nature Reviews Physics 1, 675 (2019)

    2019

  7. [15]

    Trulsen, H

    K. Trulsen, H. Zeng, and O. Gramstad, Laboratory evidence of freak waves provoked by non-uniform bathymetry, Physics of Fluids 24 (2012)

    2012

  8. [16]

    Viotti and F

    C. Viotti and F. Dias, Extreme waves induced by strong depth transitions: Fully nonlinear results, Physics of Fluids 26, 051705 (2014)

    2014

  9. [17]

    A. J. Majda, M. Moore, and D. Qi, Statistical dynamical model to predict extreme events and anomalous features in shallow water waves with abrupt depth change, Proceedings of the National Academy of Sciences 116, 3982 (2019)

    2019

  10. [18]

    Qi and E

    D. Qi and E. Vanden-Eijnden, Anomalous statistics and large deviations of turbulent water waves past a step, AIP Advances 12, 025016 (2022)

    2022

  11. [19]

    Qi and A

    D. Qi and A. J. Majda, Using machine learning to predict extreme events in complex systems, Proceedings of the National Academy of Sciences 117, 52 (2020)

    2020

  12. [20]

    J. G. Herterich and F. Dias, Extreme long waves over a varying bathymetry, Journal of Fluid Mechanics 878, 481 (2019)

    2019

  13. [21]

    Trulsen, A

    K. Trulsen, A. Raustøl, S. Jorde, and L. B. Rye, Extreme wave statistics of long-crested irregular waves over a shoal, Journal of Fluid Mechanics 882 (2020). 18

    2020

  14. [22]

    N. J. Moore, C. T. Bolles, A. J. Majda, and D. Qi, Anomalous waves triggered by abrupt depth changes: laboratory experiments and truncated KdV statistical mechanics, Journal of Nonlinear Science 30, 3235 (2020)

    2020

  15. [23]

    Zheng, Z

    Y. Zheng, Z. Lin, Y. Li, T. Adcock, Y. Li, and T. van den Bremer, Fully nonlinear simulations of unidirectional extreme waves provoked by strong depth transitions: The effect of slope, Physical Review Fluids 5, 064804 (2020)

    2020

  16. [24]

    M. S. Afzal and L. Kumar, Propagation of waves over a rugged topography, Journal of Ocean Engineering and Science (2021)

    2021

  17. [25]

    Y. Li, Y. Zheng, Z. Lin, T. A. Adcock, and T. S. van den Bremer, Surface wavepackets subject to an abrupt depth change. part 1. second-order theory, Journal of Fluid Mechanics 915 (2021)

    2021

  18. [26]

    Y. Li, S. Draycott, T. A. Adcock, and T. S. van den Bremer, Surface wavepackets subject to an abrupt depth change. part 2. experimental analysis, Journal of Fluid Mechanics 915 (2021)

    2021

  19. [27]

    Y. Li, S. Draycott, Y. Zheng, Z. Lin, T. A. Adcock, and T. S. Van Den Bremer, Why rogue waves occur atop abrupt depth transitions, Journal of Fluid Mechanics 919 (2021)

    2021

  20. [28]

    Lawrence, K

    C. Lawrence, K. Trulsen, and O. Gramstad, Statistical properties of wave kinematics in long-crested irregular waves propagating over non-uniform bathymetry, Physics of Fluids 33, 046601 (2021)

    2021

  21. [29]

    Mendes, A

    S. Mendes, A. Scotti, M. Brunetti, and J. Kasparian, Non-homogeneous analysis of rogue wave probability evolution over a shoal, Journal of Fluid Mechanics 939 (2022)

    2022

  22. [30]

    Crowdy, Solving problems in multiply connected domains (SIAM, 2020)

    D. Crowdy, Solving problems in multiply connected domains (SIAM, 2020)

    2020

  23. [31]

    P. D. Lax, Periodic solutions of the KdV equation, Communications on Pure and Applied Mathematics 28, 141 (1975)

    1975

  24. [32]

    R. S. Johnson, A Modern Introduction to the Mathematical Theory of Water Waves, Vol. 19 (Cambridge university press, 1997)

    1997

  25. [33]

    G. B. Whitham, Linear and nonlinear waves , Vol. 42 (John Wiley & Sons, 2011)

    2011

  26. [34]

    Sun and N

    H. Sun and N. J. Moore, On normal and non-normal wave statistics implied by a canonical–microcanonical Gibbs ensemble of the truncated KdV system, Journal of Statistical Physics 190, 4 (2023)

    2023

  27. [35]

    D. M. Ambrose, R. Camassa, J. L. Marzuola, R. M. McLaughlin, Q. Robinson, and J. Wilkening, Numerical algorithms for water waves with background flow over obstacles and topography, Advances in Computational Mathematics 48, 46 (2022)

    2022

  28. [36]

    Ross, A First Course in Probability: 8th edition (Pearson Education, 2010)

    S. Ross, A First Course in Probability: 8th edition (Pearson Education, 2010)

    2010

  29. [37]

    P. J. Baddoo, N. J. Moore, A. U. Oza, and D. G. Crowdy, Generalization of waving-plate theory to multiple interacting swimmers, Communications on Pure and Applied Mathematics 76, 3811 (2023)

    2023

  30. [38]

    Bajars, J

    J. Bajars, J. E. Frank, and B. J. Leimkuhler, Weakly coupled heat bath models for gibbs-like invariant states in nonlinear wave equations, Nonlinearity 26 (2013)

    2013

  31. [39]

    We find that the numerical optimization typically converges in only a few iterations, confirming that Eq

    in order to determine the rejection constant M . We find that the numerical optimization typically converges in only a few iterations, confirming that Eq. (47) provides an effective starting guess. B. The rejection algorithm All of the pieces are now in place to implement the ...

  32. [40]

    H. Sun, N. J. Moore, and F. Bao, Parameter estimation for the truncated KdV model through a direct filter method, Journal of Machine Learning for Modeling and Computing 4 (2023)

    2023

  33. [41]

    Majda and X

    A. Majda and X. Wang, Nonlinear Dynamics and Statistical Theories for Basic Geophysical Flows (Cambridge University Press, 2006)

    2006

  34. [42]

    R. V. Abramov, G. Kovaˇ ciˇ c, and A. J. Majda, Hamiltonian structure and statistically relevant conserved quan- tities for the truncated burgers-hopf equation, Communications on Pure and Applied Mathematics: A Journal Issued by the Courant Institute of Mathematical Sciences 5...

    2003

  35. [43]

    Skare, E

    Ø. Skare, E. Bølviken, and L. Holden, Improved sampling-importance resampling and reduced bias importance sampling, Scandinavian Journal of Statistics 30, 719 (2003)

    2003

  36. [44]

    W. A. Strauss, Partial differential equations: An introduction (John Wiley & Sons, 2007)

    2007

  37. [45]

    J. C. Lagarias, J. A. Reeds, M. H. Wright, and P. E. Wright, Convergence properties of the Nelder–Mead simplex method in low dimensions, SIAM Journal on optimization 9, 112 (1998)

    1998

  38. [46]

    Zhang, P

    Z. Zhang, P. Jaiswal, and R. Rai, Featurenet: Machining feature recognition based on 3d convolution neural network, Computer-Aided Design 101, 12 (2018)

    2018

  39. [47]

    D. M. Pelt and J. A. Sethian, A mixed-scale dense convolutional neural network for image analysis, Proceedings of the National Academy of Sciences 115, 254 (2018)

    2018

  40. [48]

    Qi, Unambiguous models and machine learning strategies for anomalous extreme events in turbulent dynamical system, Entropy 26, 522 (2024)

    D. Qi, Unambiguous models and machine learning strategies for anomalous extreme events in turbulent dynamical system, Entropy 26, 522 (2024)

    2024

  41. [49]

    Pedlosky, Geophysical fluid dynamics (Springer Science & Business Media, 2013)

    J. Pedlosky, Geophysical fluid dynamics (Springer Science & Business Media, 2013)

    2013

  42. [50]

    H. A. Dijkstra and M. Ghil, Low-frequency variability of the large-scale ocean circulation: A dynamical systems approach, Reviews of Geophysics 43 (2005)

    2005

  43. [51]

    N. J. Moore and J. Mac Huang, Large-scale circulation reversals explained by pendulum correspondence, Journal of Fluid Mechanics 993, A3 (2024)

    2024

  44. [52]

    J. A. Whitehead, Energy Flow and Earth (Springer, 2024)

    2024

  45. [53]

    B. R. Sutherland, U. Achatz, C.-c. P. Caulfield, and J. M. Klymak, Recent progress in modeling imbalance in the atmosphere and ocean, Physical Review Fluids 4, 010501 (2019). 19

    2019

  46. [54]

    G. P. Chini, G. Michel, K. Julien, C. B. Rocha, and P. C. Colm-cille, Exploiting self-organized criticality in strongly stratified turbulence, Journal of Fluid Mechanics 933, A22 (2022)

    2022

  47. [55]

    Camassa, R

    R. Camassa, R. M. McLaughlin, M. N. Moore, and K. Yu, Stratified flows with vertical layering of density: experimental and theoretical study of flow configurations and their stability, Journal of fluid mechanics 690, 571 (2012)

    2012

  48. [56]

    Mac Huang and N

    J. Mac Huang and N. J. Moore, Morphological attractors in natural convective dissolution, Physical Review Letters 128, 024501 (2022)

    2022

  49. [57]

    D. Han, Q. Wang, and X. Wang, Dynamic transitions and bifurcations for thermal convection in the superposed free flow and porous media, Physica D: Nonlinear Phenomena 414, 132687 (2020)

    2020

  50. [58]

    S.-H. Chiu, M. N. J. Moore, and B. Quaife, Viscous transport in eroding porous media, Journal of Fluid Mechanics 893, A3 (2020)

    2020

  51. [59]

    McCurdy, N

    M. McCurdy, N. J. Moore, and X. Wang, Predicting convection configurations in coupled fluid–porous systems, Journal of Fluid Mechanics 953, A23 (2022)

    2022

  52. [60]

    N. J. Moore, J. Cherry, S.-H. Chiu, and B. D. Quaife, How fluid-mechanical erosion creates anisotropic porous media, Physica D: Nonlinear Phenomena 445, 133634 (2023). Appendix A: V alidation with two-mode exact solutions In the case K = 2, it is possible to solve exactly for ...

    2023

  53. [61]

    (A2) Therefore, the moduli can then be written as (R1, R2) = r E0 2π (cos(ϕ), sin(ϕ)) , (A3) for some ϕ ∈ [0, 2π)

    = E0. (A2) Therefore, the moduli can then be written as (R1, R2) = r E0 2π (cos(ϕ), sin(ϕ)) , (A3) for some ϕ ∈ [0, 2π). Thus, to select a set of arbitrary modes ˆ u1 and ˆu2 satisfying the energy constraint, it suffices to specify the values θ1, θ2, ϕ∈ [0, 2π). For example, t...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.