Pith. sign in

REVIEW 3 major objections 5 minor 35 references

Evaluation of Neural Surrogates for Physical Modelling Synthesis of Nonlinear Elastic Plates

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper finds that neural surrogates trained on short sequences fail to capture the long-sequence dynamics of nonlinear elastic plates, and shows why time-domain error alone is a misleading evaluation metric.

desk verdict Useful negative-result benchmark for neural plate surrogates, but the spectral 'excess' diagnosis rests on a backwards Gaussian-width argument and needs recomputation against ground truth. read the letter →

arxiv 2507.12563 v1 pith:5F2MEWYR submitted 2025-07-16 cs.SD cs.LGeess.AS

Carlos De La Vega Martin , Rodrigo Diaz Fernandez , Mark Sandler This is my paper
classification cs.SDcs.LGeess.AS
keywords physicalmodellingsynthesisnonlinearelasticplatesBergerplatemodelneuralsurrogateslongsequencepredictionspectraldiagnosticsKoopmanoperatorstatespacemodels
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 tests whether neural networks can replace numerical solvers for synthesizing drum-like plate sounds, by training six architectures on short blocks and rolling them out autoregressively. It finds that although some models look good on single-block prediction, none captures the plate dynamics over long sequences. The key diagnostic is spectral: models inject excess energy at high spatial wavenumbers, and the Fourier Neural Operator decouples spatial from temporal frequency behaviour. If the result holds, current neural surrogates are not suitable for real-time physical-modelling audio synthesis, and evaluation practice must include spectral diagnostics.

What carries the argument

The load-bearing machinery is the Berger plate model, a nonlinear PDE for thin elastic plates, projected onto the plate's linear eigenfunctions so that the nonlinearity appears as a global tension modulation of otherwise linear modal oscillators. Ground-truth trajectories are generated with the Functional Transformation Method (FTM), and models are trained with MSE loss using short output blocks, temporal bundling, and the pushforward trick. The evaluation compares normalized time-domain errors but also uses spectrograms and time-averaged radial spatial power spectra to reveal where energy accumulates in space and time. The Koopman-inspired LTI models use autoencoders with diagonal linear latent dynamics, whose eigenvalues are clipped to remain stable.

What would settle it

Compute the time-averaged radial spatial power spectrum of the ground-truth FTM trajectories themselves for the stated 0.02 to 0.1 m Gaussian initial widths, and compare the energy above 50 $m^{-1}$ in ground truth and model outputs; if the ground truth carries non-negligible energy in that band, the excess-energy diagnosis collapses. A simpler check is whether any model, when rolled out for a full second, conserves total modal energy or grows without bound within a few hundred steps.

Watch

Extended reading notes

Core claim

The paper's central claim is stated in its conclusions: "none of the models are able to capture the dynamics of the plate in the long sequence prediction task, making them not suitable in the current state." Even the best time-domain model, the Linear Recurrent Unit, degrades badly in autoregressive rollout, and training with the pushforward trick does not meaningfully help. The paper shows that looking only at prediction error in the time domain is insufficient, because the models tend to add energy higher in the spectrum, especially beyond roughly 50 $m^{-1}$ where the Gaussian initial conditions imply little energy should exist. In the FNO case, the spatial spectrum looks similar across sequence lengths while the temporal spectrum becomes unstable, suggesting the model is not learning coherent spatiotemporal modes. This decorrelation is linked to a conceptual issue: Koopman-inspired autoencoder models compress into a low-dimensional latent space rather than expanding into the high-dimensional space that Koopman theory would suggest.

Load-bearing premise

The claim that models add excess high-wavenumber energy rests on the assumption that the true plate response has almost no energy above roughly 50 $m^{-1}$, inferred only from the minimum 0.02 m width of the Gaussian initial conditions, and this expected floor is never computed from the actual ground-truth power spectra.

Editorial extensions

If this is right

  • None of the evaluated models, trained and rolled out this way, is suitable as a stand-in for numerical solvers in nonlinear plate audio synthesis.
  • Single-block prediction quality does not transfer to autoregressive rollout, so reported time-domain error on short blocks should not be used as the sole success metric.
  • The excess energy concentrated above roughly 50 m^-1 means neural surrogates colour the sound with spurious high-frequency content even when overall error looks low.
  • Pushforward training and temporal bundling, as applied here, do not meaningfully stabilise long rollouts.
  • Spatial and temporal frequency content must be evaluated together, because spatial agreement can coexist with temporal instability, as the FNO results show.

Reading between the lines

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

  • A direct test of a spectral-loss training variant, such as multi-resolution spectrogram loss, could reveal whether the high-wavenumber excess is an artifact of MSE training or an architectural limit.
  • Because the ground truth is generated from Gaussian velocity profiles, percussive or impulse-like excitations with broadband spatial content may change the failure pattern; evaluating on such inputs is a natural next experiment.
  • The plate's dispersion relation, linking spatial and temporal frequencies, could serve as a physics-based metric: a surrogate that has truly learned the modes should trace the same dispersion curve as the FTM solution.
  • The diagnostic recipe implied by the paper, reporting energy in physical bands above the excitation floor, could be applied to other learned physical-modelling synthesis methods.
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 evaluates seven neural surrogates for the Berger nonlinear plate model used in physical modelling synthesis: a Fourier Neural Operator, three Koopman/LTI variants, S5, and LRU with and without the pushforward trick. Each model is trained on short temporal blocks and then evaluated both on single-block prediction and on autoregressive long-sequence rollout. The main empirical claims are that single-block prediction errors do not predict rollout quality, that LRU is the strongest single-block model but still degrades substantially in rollout, and that spectral analysis reveals an excess of high-wavenumber spatial energy and a decoupling between spatial and temporal spectral behavior in the model outputs. The paper concludes that current neural surrogates are not suitable for long-sequence audio synthesis and that spectral diagnostics are needed in evaluation.

Significance. If the findings hold, the paper provides a useful negative benchmark for neural PDE surrogates in the audio physical-modelling setting. The study is honest in reporting failure rather than overclaiming success, and it addresses a real gap in evaluation practice by going beyond scalar time-domain errors and examining spectral content. The normalized metrics, the comparison across several model families, and the three-seed reporting are strengths, as is the availability of the architecture code. The main weakness is that the spectral 'excess energy' conclusion rests on an unsupported ground-truth spectral floor, and the strong universal negative claim is based on a small test set. The central rollout-error finding is credible, but the spectral evidence needs to be verified against the actual FTM ground-truth spectra before the paper's main evaluation recommendation is fully established.

major comments (3)
  1. [Section 4, Fig. 3] The claim that 'we would not expect to see much energy' beyond wavenumber 50 m^-1 is not supported by the stated initial-condition parameters. For a Gaussian velocity profile with spatial standard deviation sigma, the wavenumber power spectrum decays as exp(-sigma^2 k^2); the minimum sigma of 0.02 m yields the broadest spectrum, with only -4.3 dB at k = 50 m^-1 and -8.5 dB at k = 70 m^-1 relative to the peak. The paper should compute the actual radial power spectrum of the FTM ground truth, quantify the model excess in specific wavenumber bands, and revise the 'models add high-wavenumber energy' claim accordingly. This is load-bearing because the spectral diagnosis is a central motivation for the paper's main recommendation that spectral evaluation is necessary.
  2. [Section 4 and Appendix C.4] The conclusion that 'none of the models are able to capture the dynamics of the plate in the long sequence prediction task' is a strong universal negative statement supported by a test set of only 10 trajectories and three seeds. The paper should report the distribution of rollout errors across the test trajectories and seeds, show confidence intervals, and frame the negative conclusion as applying to the evaluated benchmark rather than to the entire class of neural surrogates. This does not require new experiments, but the current wording overgeneralizes from a small sample.
  3. [Section 4, Fig. 2 and Fig. 3] The paper describes the long-sequence task as generating up to 4000 steps, which at a 16 kHz sampling rate is only 250 ms and is far shorter than the 'hundreds of thousands of timesteps' mentioned in the introduction. The text also refers to spectrograms 'for the full second (16000 samples)', while Figure 3 is described as averaged over the first 4000 samples. Please clarify exactly which rollout lengths are used in each figure and, if possible, report how the errors evolve over the full 1 s trajectory. Without this clarification, the link between the reported rollout failure and audio-scale synthesis remains unclear.
minor comments (5)
  1. [Section 1 and Section 5] There are several typographical errors: 'This is is particularly relevant' in Section 1, 'neccesarily' and 'woud' in Section 5, and 'as see in Fig. 2b' in Section 4. These should be fixed in a revision.
  2. [Figure 1 caption] The caption says 'output legth 49'; this should be 'output length 49'.
  3. [Section 4] The sentence 'the last step of the output is used as input for the next prediction of the next block' is redundant; 'the last step of the output is used as input for the next block' would be clearer.
  4. [Figure 3] The legend and line styles for the ground-truth curve should be made more visually distinct, for example with a thicker black line, to make the comparison easier for readers with color vision deficiencies.
  5. [Appendix A, Eq. (8)] The symbols S(u) and S0 are used in the nonlinear tension term but S0 is not explicitly defined in the main text or appendix; please define it at first use.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the benchmark compares neural surrogates against an external FTM solver ground truth on held-out trajectories, and no fitted parameter is repackaged as a prediction.

full rationale

This paper is an empirical benchmark study, not a derivation. The ground-truth trajectories are generated by the Functional Transformation Method (FTM), an external numerical solver, and the neural models are trained on short blocks and then evaluated autoregressively on held-out trajectories. This is the intended design for a surrogate-accuracy study. The central claim that models fail at long-sequence prediction is based on measured errors against this external ground truth, not on any quantity fitted from the target results. The models are fitted to solver-generated training data, but the evaluation is on separate test trajectories, and the paper does not rename a fitted constant as a predicted quantity. The spectral analysis in Section 4 relies on a qualitative expectation about high-wavenumber energy based on the Gaussian initial-condition width; while this expectation is not computed from the actual ground-truth spectra and may be directionally questionable, that is a weakness in the supporting evidence, not circular reasoning. The paper's self-citations (De La Vega Martin & Sandler 2023; Diaz et al. 2024) are used as related-work references for known error accumulation in neural PDE surrogates; they are not load-bearing for the present results. No equation in the paper reduces to its own input, and no uniqueness theorem or ansatz is imported from the authors' prior work to force the conclusions. Therefore, the circularity score is 0.

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

The central evaluation rests on the Berger approximation, the accuracy of the FTM ground truth, and a set of hand-chosen dataset and training parameters. The nonlinear coefficient C_NL is never given, which blocks exact reproduction. No new physical or mathematical entities are introduced.

free parameters (5)
  • Berger nonlinear coupling coefficient C_NL = not stated
    The nonlinear forcing term in Eq. 10 depends on C_NL, but Appendix B gives no numeric value, so the exact dataset cannot be reproduced from the paper alone.
  • Number of FTM modes per dimension = 15 (225 total)
    Chosen for the solver projection; affects ground-truth accuracy and high-wavenumber content, and is not validated against a converged solution.
  • Initial condition parameters = velocity 5 to 25 m/s, Gaussian std 0.02 to 0.1 m
    Randomized initial-condition ranges define the dataset distribution; conclusions about model failure are conditioned on this range.
  • Temporal bundling output lengths = 49, 199, 399 steps
    Training and evaluation block lengths are hand-set experimental choices that influence the observed error accumulation and model ranking.
  • FNO truncation modes and hidden channels = 20 modes per dimension, 32 hidden channels
    Hand-chosen architecture capacity; no ablation is reported, so capacity differences may confound cross-model comparisons.
assumptions (5)
  • domain assumption Berger plate approximation: tension nonlinearity is homogeneous and depends only on the global state
    Adopted from Berger (1954) via Avanzini et al. (2012), Eq. 8 in Appendix A. Underlies all training data; if it poorly represents real drum membranes, conclusions about audio synthesis are limited.
  • standard math FTM with 225 modes yields ground truth accurate enough for model comparison
    FTM is established in the literature, but truncation at 15 modes per dimension is an approximation and its error against the full PDE is not quantified in the paper.
  • ad hoc to paper Negligible ground-truth energy beyond wavenumber 50 m^-1
    Used in Section 4 and Figure 3 to identify excess high-wavenumber energy; inferred from the minimum Gaussian std of 0.02 m but never computed or quantified from the actual ground truth.
  • domain assumption Gaussian velocity initial conditions and no external forcing are sufficient to represent audio synthesis use cases
    The dataset in Section 2.1 contains only pluck-like initial conditions; real drum excitation may include sustained forcing, which the models were never trained or tested on.
  • domain assumption MSE training and standard hyperparameters make a fair test of each architecture class
    No extensive hyperparameter search or ablations are reported; model rankings might change with different tuning, so the negative result is conditional on this setup.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Evaluation of Neural Surrogates for Physical Modelling Synthesis of Nonlinear Elastic Plates." pith.science (2026). https://pith.science/paper/5F2MEWYR

@misc{pith2026250712563,
  author       = {Pith},
  title        = {Pith review of: Evaluation of Neural Surrogates for Physical Modelling Synthesis of Nonlinear Elastic Plates},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5F2MEWYR}},
  note         = {Machine review of arXiv:2507.12563}
}
read the original abstract

Physical modelling synthesis aims to generate audio from physical simulations of vibrating structures. Thin elastic plates are a common model for drum membranes. Traditional numerical methods like finite differences and finite elements offer high accuracy but are computationally demanding, limiting their use in real-time audio applications. This paper presents a comparative analysis of neural network-based approaches for solving the vibration of nonlinear elastic plates. We evaluate several state-of-the-art models, trained on short sequences, for prediction of long sequences in an autoregressive fashion. We show some of the limitations of these models, and why is not enough to look at the prediction error in the time domain. We discuss the implications for real-time audio synthesis and propose future directions for improving neural approaches to model nonlinear vibration.

Figures

Figures reproduced from arXiv: 2507.12563 by the authors.

Figure 1
Figure 1. Relative MAE per block during the first 250 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. Radial spatial power spectrum of the predictions [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Architecture of LTI-MLP showing linear dynamics post-processing, involving an MLP using the square of [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

35 extracted references · 20 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    Efficient synthesis of tension modulation in strings and membranes based on energy estimation

    Avanzini, F., Marogna, R., and Bank, B. Efficient synthesis of tension modulation in strings and membranes based on energy estimation. The Journal of the Acoustical Society of America, 131 0 (1): 0 897--906, January 2012. ISSN 0001-4966. doi:10.1121/1.3651097. URL http://asa.scitation.org/doi/10.1121/1.3651097

    work page doi:10.1121/1.3651097 2012

  3. [3]

    Berger, H. M. A New Approach to the Analysis of Large Deflections of Plates . Doctoral Thesis , California Institute of Technology, January 1954. URL https://resolver.caltech.edu/CaltechETD:etd-12042003-163054

    work page 1954

  4. [4]

    Blelloch, G. E. Prefix sums and their applications. May 2004. doi:10.1184/R1/6608579.v1. URL https://kilthub.cmu.edu/articles/journal_contribution/Prefix_sums_and_their_applications/6608579/1. Publisher: Carnegie Mellon University

    work page arXiv 2004

  5. [5]

    Brandstetter, J., Welling, M., and Worrall, D. E. Lie Point Symmetry Data Augmentation for Neural PDE Solvers . In Proceedings of the 39th International Conference on Machine Learning , pp.\ 2241--2256. PMLR, June 2022 a . URL https://proceedings.mlr.press/v162/brandstetter22a.html. ISSN: 2640-3498

    work page 2022

  6. [6]

    E., and Welling, M

    Brandstetter, J., Worrall, D. E., and Welling, M. Message Passing Neural PDE Solvers . In Proceedings of ICLR 2022 , April 2022 b . URL https://openreview.net/forum?id=vSix3HPYKSU

    work page 2022

  7. [7]

    L., Brunton, B

    Brunton, S. L., Brunton, B. W., Proctor, J. L., and Kutz, J. N. Koopman Invariant Subspaces and Finite Linear Representations of Nonlinear Dynamical Systems for Control . PLOS ONE, 11 0 (2): 0 e0150171, February 2016. ISSN 1932-6203. doi:10.1371/journal.pone.0150171. URL https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0150171. Publisher:...

    doi:10.1371/journal.pone.0150171 2016

  8. [8]

    S., Giampaolo, F., Rozza, G., Raissi, M., and Piccialli, F

    Cuomo, S., di Cola, V. S., Giampaolo, F., Rozza, G., Raissi, M., and Piccialli, F. Scientific machine learning through physics-informed neural networks: where we are and what's next, February 2022. URL http://arxiv.org/abs/2201.05624. arXiv:2201.05624 [physics] type: article

    arXiv 2022

Show all 35 references
  1. [9]

    and Sandler, M

    De La Vega Martin, C. and Sandler, M. Physical Modelling of Stiff Membrane Vibration using Neural Networks with Spectral Convolution Layers . In Proceedings of the 10th Convention of the European Acoustics Association , Turin, 2023

    2023

  2. [10]

    Towards Efficient Modelling of String Dynamics : A Comparison of State Space and Koopman based Deep Learning Methods

    Diaz, R., De La Vega Martin, C., and Sandler, M. Towards Efficient Modelling of String Dynamics : A Comparison of State Space and Koopman based Deep Learning Methods . In Proc. Int . Conf . Digital Audio Effects ( DAFx -24) , 2024. URL https://dafx24.surrey.ac.uk/wp-content/up...

    2024

  3. [11]

    Fletcher, N. H. and Rossing, T. D. The physics of musical instruments. Springer New York, New York, NY, 1st edition edition, January 1991. ISBN 978-0-387-94151-6. doi:10.1007/978-1-4612-2980-3. URL http://link.springer.com/10.1007/978-1-4612-2980-3

    1991 doi

  4. [12]

    Deep learning

    Goodfellow, I., Bengio, Y., and Courville, A. Deep learning. MIT Press, 2016

    2016

  5. [13]

    Modeling sequences with structured state spaces

    Gu, A. Modeling sequences with structured state spaces. PhD thesis, Stanford University, Stanford, California, 2023

    2023

  6. [14]

    Efficiently Modeling Long Sequences with Structured State Spaces , August 2022

    Gu, A., Goel, K., and Ré, C. Efficiently Modeling Long Sequences with Structured State Spaces , August 2022. URL http://arxiv.org/abs/2111.00396. arXiv:2111.00396 [cs]

    2022 arXiv

  7. [15]

    Diagonal State Spaces are as Effective as Structured State Spaces

    Gupta, A., Gu, A., and Berant, J. Diagonal State Spaces are as Effective as Structured State Spaces . In Advances in Neural Information Processing Systems , volume 35, May 2022. URL https://openreview.net/forum?id=RjS0j6tsSrf

    2022

  8. [16]

    A., Shen, Q., Kawaguchi, K., and Karniadakis, G

    Hu, Z., Daryakenari, N. A., Shen, Q., Kawaguchi, K., and Karniadakis, G. E. State-space models are accurate and efficient neural operators for dynamical systems, September 2024. URL http://arxiv.org/abs/2409.03231. arXiv:2409.03231

    2024 arXiv

  9. [17]

    Huhtala, V., Juvela, L., and Schlecht, S. J. KLANN : Linearising Long - Term Dynamics in Nonlinear Audio Effects Using Koopman Networks . IEEE Signal Processing Letters, 31: 0 1169--1173, 2024. ISSN 1558-2361. doi:10.1109/LSP.2024.3389465. URL https://ieeexplore.ieee.org/docum...

    2024

  10. [18]

    Benchmarking Autoregressive Conditional Diffusion Models for Turbulent Flow Simulation , January 2024

    Kohl, G., Chen, L.-W., and Thuerey, N. Benchmarking Autoregressive Conditional Diffusion Models for Turbulent Flow Simulation , January 2024. URL http://arxiv.org/abs/2309.01745. arXiv:2309.01745 [physics]

    2024 arXiv

  11. [19]

    Koopman, B. O. Hamiltonian Systems and Transformation in Hilbert Space . Proceedings of the National Academy of Sciences, 17 0 (5): 0 315--318, May 1931. doi:10.1073/pnas.17.5.315. URL https://www.pnas.org/doi/10.1073/pnas.17.5.315. Publisher: Proceedings of the National Acade...

    1931 doi

  12. [20]

    Neural Operator : Learning Maps Between Function Spaces With Applications to PDEs

    Kovachki, N., Li, Z., Liu, B., Azizzadenesheli, K., Bhattacharya, K., Stuart, A., and Anandkumar, A. Neural Operator : Learning Maps Between Function Spaces With Applications to PDEs . Journal of Machine Learning Research, 24 0 (1): 0 4061--4157, January 2023. ISSN 1533-7928. ...

    2023

  13. [21]

    Kramer, M. A. Nonlinear principal component analysis using autoassociative neural networks. AIChE Journal, 37 0 (2): 0 233--243, 1991. ISSN 1547-5905. doi:10.1002/aic.690370209. URL https://onlinelibrary.wiley.com/doi/abs/10.1002/aic.690370209. \_eprint: https://onlinelibrary....

    1991 doi

  14. [22]

    Lu, L., Jin, P., Pang, G., Zhang, Z., and Karniadakis, G. E. Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators. Nature Machine Intelligence, 3 0 (3): 0 218--229, March 2021. ISSN 2522-5839. doi:10.1038/s42256-021-00302-5. URL h...

    2021 doi

  15. [23]

    N., and Brunton, S

    Lusch, B., Kutz, J. N., and Brunton, S. L. Deep learning for universal linear embeddings of nonlinear dynamics. Nature Communications, 9 0 (1): 0 4950, November 2018. ISSN 2041-1723. doi:10.1038/s41467-018-07210-0. URL https://www.nature.com/articles/s41467-018-07210-0. Number...

    2018 doi

  16. [24]

    Spectrum of the Koopman Operator , Spectral Expansions in Functional Spaces , and State - Space Geometry

    Mezić, I. Spectrum of the Koopman Operator , Spectral Expansions in Functional Spaces , and State - Space Geometry . Journal of Nonlinear Science, 30 0 (5): 0 2091--2145, October 2020. ISSN 1432-1467. doi:10.1007/s00332-019-09598-5. URL https://doi.org/10.1007/s00332-019-09598-5

    2020 doi

  17. [25]

    Koopman Operator , Geometry , and Learning of Dynamical Systems

    Mezić, I. Koopman Operator , Geometry , and Learning of Dynamical Systems . Notices of the American Mathematical Society, 68 0 (07): 0 1, August 2021. ISSN 0002-9920, 1088-9477. doi:10.1090/noti2306. URL https://www.ams.org/notices/202107/rnoti-p1087.pdf

    2021 doi

  18. [26]

    E., and Riemer-Sørensen, S

    Michałowska, K., Goswami, S., Karniadakis, G. E., and Riemer-Sørensen, S. Neural Operator Learning for Long - Time Integration in Dynamical Systems with Recurrent Neural Networks . In 2024 International Joint Conference on Neural Networks ( IJCNN ) , pp.\ 1--8, Yokohama, Japan...

    2024

  19. [27]

    Morse, P. M. and Ingard, K. U. Theoretical acoustics. International Series In Pure And Applied Physics . McGraw-Hill, 1st edition edition, 1968

    1968

  20. [28]

    B., Ren, P., Mahoney, M

    Naiman, I., Erichson, N. B., Ren, P., Mahoney, M. W., and Azencot, O. Generative Modeling of Regular and Irregular Time Series Data via Koopman VAEs . October 2023. URL https://openreview.net/forum?id=eY7sLb0dVF

    2023

  21. [29]

    L., Gu, A., Fernando, A., Gulcehre, C., Pascanu, R., and De, S

    Orvieto, A., Smith, S. L., Gu, A., Fernando, A., Gulcehre, C., Pascanu, R., and De, S. Resurrecting recurrent neural networks for long sequences. In Proceedings of the 40th International Conference on Machine Learning , volume 202 of ICML '23 , pp.\ 26670--26698, Honolulu, Haw...

    2023

  22. [30]

    D., Schlecht, S

    Parker, J. D., Schlecht, S. J., Rabenstein, R., and Schäfer, M. Physical Modeling using Recurrent Neural Networks with Fast Convolutional Layers . In Proceedings of the 25th International Conference on Digital Audio Effects ( DAFx20in22 ) , Vienna, Austria, June 2022. doi:10.4...

    2022 doi arXiv

  23. [31]

    FourCastNet : A Global Data -driven High -resolution Weather Model using Adaptive Fourier Neural Operators , February 2022

    Pathak, J., Subramanian, S., Harrington, P., Raja, S., Chattopadhyay, A., Mardani, M., Kurth, T., Hall, D., Li, Z., Azizzadenesheli, K., Hassanzadeh, P., Kashinath, K., and Anandkumar, A. FourCastNet : A Global Data -driven High -resolution Weather Model using Adaptive Fourier...

    2022 arXiv

  24. [32]

    Smith, J. T. H., Warrington, A., and Linderman, S. Simplified State Space Layers for Sequence Modeling . In Proceedings of ICLR 2023 , Kigali, Rwanda, May 2023. URL https://openreview.net/forum?id=Ai8Hw3AXqks

    2023

  25. [33]

    and Rabenstein, R

    Trautmann, L. and Rabenstein, R. Digital Sound Synthesis by Physical Modeling Using the Functional Transformation Method . Springer US, Boston, MA, 2003. ISBN 978-1-4613-4900-6 978-1-4615-0049-0. doi:10.1007/978-1-4615-0049-0. URL http://link.springer.com/10.1007/978-1-4615-0049-0

    2003 doi

  26. [34]

    Koopman neural operator as a mesh-free solver of non-linear partial differential equations

    Xiong, W., Huang, X., Zhang, Z., Deng, R., Sun, P., and Tian, Y. Koopman neural operator as a mesh-free solver of non-linear partial differential equations. Journal of Computational Physics, 513: 0 113194, September 2024. ISSN 0021-9991. doi:10.1016/j.jcp.2024.113194. URL http...

    2024

  27. [35]

    Continuous PDE Dynamics Forecasting with Implicit Neural Representations

    Yin, Y., Kirchmeyer, M., Franceschi, J.-Y., Rakotomamonjy, A., and Gallinari, P. Continuous PDE Dynamics Forecasting with Implicit Neural Representations . In Proceedings of ICLR 2023 , Kigali, Rwanda, May 2023. URL https://openreview.net/forum?id=B73niNjbPs

    2023

Pith tools

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