Pith. sign in

REVIEW 4 major objections 6 minor 39 references

An effective physics-informed neural operator framework for predicting wavefields

T0 review · 4 major / 6 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 Adding the Helmholtz residual to a neural operator's loss cuts wavefield error by up to 53.1%.

desk verdict A modest but well-controlled PINO paper whose central causal claim is undercut by an unaddressed singularity in the PDE loss. read the letter →

arxiv 2507.16431 v1 pith:3OLBVCV7 submitted 2025-07-22 physics.geo-ph cs.LG

Xiao Ma , Tariq Alkhalifah This is my paper
classification physics.geo-phcs.LG MSC 68T0765N0686A15 PACS 91.30.-f02.60.-x07.05.Mh
keywords physics-informedneuraloperatorHelmholtzequationseismicwavefieldpredictionconvolutionalscatteredgeneralizationfullwaveforminversionfrequency-domain
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

This paper tries to establish that a convolutional neural operator can learn the frequency-domain Helmholtz solution operator from a small training set if the Helmholtz residual is added to the training loss, because the physics constraint acts as a regularizer. The network maps the analytically computed background wavefield and the velocity model to the scattered wavefield, and the physics loss penalizes violations of the scattering-form wave equation pointwise with an eighth-order finite-difference Laplacian. The headline evidence is that on held-out velocity models the relative L2 error drops from 0.50 to 0.23 at 12 Hz (a 53.1% relative reduction), with smaller gains at 8 and 10 Hz. The paper further reports that PICNO preserves wavefront detail at high frequency and on out-of-distribution fault models, while the purely data-driven CNO shows phase mismatch and amplitude damping. If the claim is right, fast wavefield surrogates for seismic imaging and waveform inversion do not necessarily require massive training datasets.

What carries the argument

The load-bearing object is the physics loss $J_{\mathrm{pde}}$: the squared residual of the scattering-form Helmholtz equation, evaluated pointwise with an eighth-order finite-difference Laplacian over the entire grid. Substituting the network output $G_\theta$ into this residual gives gradients that push the prediction toward satisfying the wave equation, not merely matching labels. The second load-bearing object is the analytically computed background wavefield $U_0$, a Hankel function of the second kind for a constant background, which encodes source position and frequency as two input channels and defines the target as the scattered field $\delta U = U - U_0$.

What would settle it

Retrain PICNO with the source neighborhood removed from the physics loss (and the same hyperparameters), then compare test errors; if the 53.1% improvement over CNO shrinks sharply, the claimed regularization is coming from the singular source region rather than from honoring the wave equation globally. A second check is to evaluate the PDE residual of PICNO's predictions on out-of-distribution models; if that residual is not systematically smaller than CNO's, the physics constraint is not being enforced where it is claimed to matter.

Watch

Extended reading notes

Core claim

The central claim is that physics-informed training changes what the operator learns, not just how well it fits. PICNO uses the same convolutional neural operator architecture as the data-driven baseline, with convolutions performed in physical space, and it takes as input the real and imaginary parts of the background wavefield $U_0$ plus the velocity model $v$, producing the real and imaginary parts of the scattered wavefield $\delta U$. The physics loss $J_{\mathrm{pde}}$ is the squared residual of the scattering-form Helmholtz equation, $\frac{\omega^2}{v^2}G_\theta + \nabla^2 G_\theta + \omega^2(\frac{1}{v^2}-\frac{1}{v_0^2})U_0$, evaluated pointwise on the whole grid with an eighth-order finite-difference stencil. The paper reports that, with 3600 training samples per frequency, PICNO achieves relative L2 errors of 0.18, 0.22, and 0.23 at 8, 10, and 12 Hz, versus 0.25, 0.34, and 0.50 for CNO, and that the gap persists when sources are randomly placed throughout the model and on an out-of-distribution fault-model test. The authors also state that the PDE term destabilizes training, producing abrupt validation fluctuations and occasional jumps in the loss.

Load-bearing premise

The load-bearing premise is that evaluating the Helmholtz residual pointwise over the whole grid—including grid points right next to the source, where the analytic background field blows up logarithmically—gives a stable and physically meaningful training signal; the paper does not state that the source region is masked or down-weighted.

Editorial extensions

If this is right

  • On the 800-sample test set, PICNO's relative L2 error is 0.18, 0.22, and 0.23 at 8, 10, and 12 Hz against CNO's 0.25, 0.34, and 0.50, a 27.4%, 35.2%, and 53.1% relative reduction.
  • PICNO predicts a frequency-domain wavefield in about 0.83 seconds versus about 3 seconds for the numerical solver, so accuracy gains do not sacrifice the speed advantage of neural operators.
  • Training on 72 base velocity models (3600 samples per frequency) is enough for PICNO to generalize to held-out velocity models, indicating that the PDE constraint reduces the data requirement.
  • PICNO also generalizes to source locations distributed throughout the domain and to an out-of-distribution velocity model with faults, where the data-only CNO shows much larger errors.
  • The physics loss introduces training instability, with abrupt validation-error fluctuations and occasional loss jumps, so the improved generalization comes with a harder optimization landscape.

Reading between the lines

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

  • The residual $J_{\mathrm{pde}}$ is evaluated pointwise over the whole grid with no stated masking or down-weighting of the source neighborhood, where $U_0$ has a logarithmic singularity; if the source region is not excluded, part of the reported gain may reflect fitting that singular structure rather than a global wave-equation constraint.
  • A clean test of the regularization story would be to retrain with the residual computed only at points at least one wavelength from the source; if the 53.1% improvement vanishes, the source-region term is doing most of the work.
  • If the effect is a genuine inductive bias, the same scattered-field-plus-PDE-loss recipe should transfer to other elliptic equations with analytic background solutions, such as the damped Helmholtz equation, where the operator only has to learn the correction field.
  • The reported 0.83-second inference time does not include the extra cost of evaluating the PDE residual during training, so the speed claim applies to inference-only use of the trained surrogate rather than to end-to-end inversion loops.
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 / 6 minor

Summary. This manuscript introduces the physics-informed convolutional neural operator (PICNO), which predicts frequency-domain scattered wavefields for the Helmholtz equation by combining a convolutional neural operator (CNO) with a PDE-residual loss. The inputs are the analytic background wavefield U0 and a velocity model; the output is the scattered wavefield δU. Experiments on synthetic datasets derived from OpenFWI at 8, 10, and 12 Hz with fixed-depth sources (Case 1), at 12 and 14 Hz with randomly distributed source locations (Case 2), and a single out-of-distribution velocity model with faults are presented. The central claim is that the physics term improves generalization, with relative L2 errors dropping from 0.50 to 0.23 at 12 Hz (a 53.1% relative reduction) and qualitatively better wavefield detail.

Significance. The problem is well-motivated for seismic imaging and inversion. The experimental design is sound in principle: the comparison is controlled (same architecture, only the physics loss term is added), the improvements are consistent across several frequencies and both source-distribution cases, and the code is made publicly available via GitHub. If the reported gains are robust, the paper would provide a valuable demonstration that physics-informed regularization can improve the generalization of neural operators in data-sparse geophysical settings. However, the current evidence is weakened by an unstated treatment of the source singularity in the PDE loss, the lack of multiple-seed statistics, and the anecdotal nature of the out-of-distribution evaluation. These are load-bearing issues, but they are addressable within the scope of a revision.

major comments (4)
  1. [2.3, Eq. (8)] The PDE residual in Eq. (8) is stated to be evaluated pointwise over the input grid using an eighth-order finite-difference stencil, and the background field U0 in Eq. (3) is the Hankel function H0^(2), which is logarithmically singular at the source. The residual therefore includes the singular term ω^2(1/v^2 − 1/v0^2)U0 plus the Laplacian of the predicted field, and the paper never states that the source region is masked, tapered, or excluded. Because the CNO output is a smooth, band-limited convolutional field (Section 2.2), the residual near the source cannot be made small, and the gradient of Jpde is plausibly dominated by a few grid points. This makes the physics loss an uncontrolled source-dependent penalty rather than a faithful Helmholtz constraint, and it weakens the causal interpretation of the improvements in Table 1. Please specify whether a source mask or taper is used, and if not, either introduce one, modify the loss, or provide evidence that the source-region contribution does not dominate the residual.
  2. [3, Table 1] The headline results in Table 1 are single runs without variance information. The authors themselves document abrupt fluctuations and outliers in the validation error (Section 3.1, Figure 3(d)–(f)) and a sensitive dependence on the loss weight λ (Section 4.1). Without multiple random seeds or equivalent statistical treatment, the reported relative improvements of 27.4%, 35.2%, and 53.1% are not established as reproducible. Please report mean and standard deviation over at least three independent training runs (ideally five) for both CNO and PICNO at each frequency and case, and state the exact λ values used for every configuration. If the advantage does not persist across seeds, the claims should be qualified accordingly.
  3. [3.3, Figure 10] The out-of-distribution evaluation is based on a single visual example (Figure 10) with no quantitative error metric. To substantiate the claim that PICNO generalizes to faulted and sharp-discontinuity models, report the relative L2 error (or a similar metric) for the OOD model for both CNO and PICNO, and ideally for a small set of OOD models. Without numbers, the OOD conclusion is anecdotal and does not support the strong statement in the abstract and conclusions.
  4. [4.1, Figure 11] The λ-sensitivity experiment is described only through the figure caption, which labels 'PICNO case1' and 'PICNO case2' with loss ratios of 10:1 and 8:1, but the text does not specify which case or frequency this refers to, nor the exact λ values used in the main experiments. Since the authors state that λ is selected empirically and that no universal ratio exists, the chosen λ for each training configuration (Cases 1 and 2, each frequency) must be reported, and the authors should justify that the headline results are not a product of selecting λ on the validation set. Ideally, also provide a sensitivity analysis over a range of λ values for at least one representative configuration.
minor comments (6)
  1. [2.3] The section title contains a typo: 'F ormulation' should be 'Formulation'.
  2. [Eq. (3)] Check the argument of the Hankel function in Eq. (3); the wavenumber should be ω/v0 rather than ω/v0^2 as printed. If this is a typesetting error, correct it; if it reflects the implementation, the physics loss would use an incorrect background wavefield.
  3. [Figures 4–10] The captions contain garbled path strings such as '/gid00051/gid00068/...'; these are likely LaTeX artifacts and should be removed.
  4. [4] The opening sentence of Section 4 refers to 'the Example section'; the intended term is 'Results' section.
  5. [2.2] The architecture description could be clearer; in particular, 'U-Net-style patching' and the exact composition of the D, H, I, and R blocks are hard to follow. A pseudocode or a table summarizing layer configurations would aid reproducibility.
  6. [3.1] The sentence 'We show the the relative error' contains a duplicated definite article.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the PICNO test-set gains are measured against independent held-out velocity models, and the physics regularizer is not fitted to those targets.

full rationale

The central quantitative claim is Table 1, which compares relative L2 error on 800 test samples from velocity models explicitly excluded from training and validation. The physics loss Jpde (Eq. 8) is the residual of the same scattering Helmholtz equation (Eq. 2) used to generate the labels, so a network that fits the labels will also lower Jpde; this is the intended inductive bias, not a fitted parameter renamed as a prediction. The comparison is to a CNO baseline trained with the same data and the same U0 input, and no test quantity is used to set lambda or network weights. The only notable self-citation is the adoption of U0 as an input feature from Huang & Alkhalifah (2024/2025), which is a design inheritance, not a load-bearing proof and does not constrain the outcome. Two legitimate concerns outside circularity should be noted: the pointwise eighth-order residual in Eq. 8 includes the singular Hankel background U0 at the source without a stated mask, and the paper itself reports abrupt training fluctuations and lambda sensitivity; these bear on robustness and physical fidelity of the loss, not on whether the headline comparison reduces to its inputs.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The ledger is modest: one hand-tuned loss weight, no invented entities. The main burden is the unexamined near-source behavior of the physics loss and reliance on OpenFWI as representative of geologically diverse media.

free parameters (1)
  • PDE loss weight lambda = 10:1 and 8:1 tested; final value not uniquely specified
    lambda in Eq (9) balances J_pde against J_data; the paper says it was determined empirically through trial and error (Sec 2.3, Sec 4.1) and that no universal ratio exists, so the reported errors depend on this hand-tuned choice.
assumptions (5)
  • standard math The scattered-field reformulation (Eq 2) is the correct Helmholtz equation for delta U = U - U0 with constant background v0.
    Standard operator splitting used in Sec 2.1; correct given constant v0 and same source term.
  • domain assumption U0 computed analytically from the 2D Green's function (Eq 3) is accurate on the model grid.
    Requires constant v0 and a point source in an unbounded medium; boundary reflections of U0 are ignored in the residual.
  • domain assumption An eighth-order finite-difference stencil on the training grid accurately represents the Laplacian at 8-14 Hz.
    Stated in Sec 2.3 without convergence tests; accuracy depends on grid spacing relative to wavelength.
  • ad hoc to paper The pointwise PDE residual remains a meaningful training signal near the source, where U0 is singular.
    Eq (3) diverges logarithmically at the source; the paper does not describe source masking, and the admitted training instability may stem from this.
  • domain assumption OpenFWI velocity models plus four smoothed variants are representative enough to test generalization to faulted models.
    Sec 3.1 and 3.3; the OOD test uses one faulted model, no statistics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An effective physics-informed neural operator framework for predicting wavefields." pith.science (2026). https://pith.science/paper/3OLBVCV7

@misc{pith2026250716431,
  author       = {Pith},
  title        = {Pith review of: An effective physics-informed neural operator framework for predicting wavefields},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3OLBVCV7}},
  note         = {Machine review of arXiv:2507.16431}
}
read the original abstract

Solving the wave equation is fundamental for geophysical applications. However, numerical solutions of the Helmholtz equation face significant computational and memory challenges. Therefore, we introduce a physics-informed convolutional neural operator (PICNO) to solve the Helmholtz equation efficiently. The PICNO takes both the background wavefield corresponding to a homogeneous medium and the velocity model as input function space, generating the scattered wavefield as the output function space. Our workflow integrates PDE constraints directly into the training process, enabling the neural operator to not only fit the available data but also capture the underlying physics governing wave phenomena. PICNO allows for high-resolution reasonably accurate predictions even with limited training samples, and it demonstrates significant improvements over a purely data-driven convolutional neural operator (CNO), particularly in predicting high-frequency wavefields. These features and improvements are important for waveform inversion down the road.

Figures

Figures reproduced from arXiv: 2507.16431 by the authors.

Figure 1
Figure 1. The architecture of CNO. Different colors of the arrows represent different blocks. The input function is first transformed into a high-dimensional space through a lifting layer (P), then processed by a series of neural operator layers, and finally mapped back to its original spa￾tial dimensions via a projection operator (Q). Jdata to train the CNO. This hybrid loss effectively reduces the dependence on large quan￾t… view at source ↗
Figure 2
Figure 2. (a) and (b) show two representative velocity models selected from the training dataset, along with their corresponding four smoothed variants. These variations are used for data augmentation to improve the model’s robustness. is most obvious at 12 Hz, where high-frequency components dominate the wavefield. In this case, CNO struggles to recover small-scale features, resulting in significant struc￾tural deviations ( … view at source ↗
Figure 3
Figure 3. (a)-(c) The training loss curves for the 500 epochs of training for the Helmholtz equation. (d)-(f) Relative L2 error curves of CNO and PICNO evaluated on the validation dataset. Here, we compare the different losses from the data and PDE terms for predicting 8Hz, 10Hz, and 12Hz wavefields. els, where the source locations also vary along the depth and lateral directions. During the training stage, we keep all hyperp… view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Prediction results of frequency-domain wavefields for a test sample at 8 Hz. (a) Reference wavefield computed using a conventional numerical solver; (b) corresponding veloc￾ity model; (c) wavefield predicted by CNO; (d) difference between the CNO prediction and the ref…
Figure 5
Figure 5. Figure 5: The frequency domain wavefield prediction results for an instance from the test set (10Hz). across the domain, preserving both the amplitude and phase structures even under high￾frequency, random source location conditions. This indicates that the inclusion of physics …
Figure 6
Figure 6. Figure 6: The frequency domain wavefield prediction results for an instance from the test set (12Hz). (a) (b) (c) (d) [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: (a) and (b) show the training loss curves for the 800 epochs of training for the Helmholtz equation. (c) and (d) show Relative L2 error curves of CNO and PICNO evaluated on the validation dataset. Here, we compare the different losses for predicting 12Hz and 14Hz wavef…
Figure 8
Figure 8. Figure 8: The frequency domain wavefield prediction results for an instance from the test set (12Hz). (a) (b) (c) (d) (e) (f)                           [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: The frequency domain wavefield prediction results for an instance from the test set (14Hz). 3.3 Out of distribution test In addition to the in-distribution velocity model tests, we further assess the gen￾eralization capability of PICNO through an out-of-distribution (O…
Figure 10
Figure 10. Figure 10: The frequency domain wavefield prediction results for an OOD instance from the test set (14Hz). equation ( [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]
Figure 11
Figure 11. Figure 11: Validation relative error curves under three different training configurations. The results compare CNO (trained with data loss only), PICNO case1 (trained with a PDE-to-data loss ratio of 10:1), and PICNO case2 (with a ratio of 8:1). This figure illustrates how the c…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

39 extracted references · 33 canonical work pages

  1. [1]

    APACrefauthors \ 2000

    alkhalifah2000acoustic APACrefauthors Alkhalifah, T. APACrefauthors \ 2000 . An acoustic wave equation for anisotropic media An acoustic wave equation for anisotropic media . Geophysics 65 4 1239--1250

    work page 2000

  2. [2]

    , Song, C

    alkhalifah2021wavefield APACrefauthors Alkhalifah, T. , Song, C. , bin Waheed, U. \ Hao, Q. APACrefauthors \ 2021 . Wavefield solutions from machine learned functions constrained by the Helmholtz equation Wavefield solutions from machine learned functions constrained by the helmholtz equation . Artificial Intelligence in Geosciences 2 11--19

    work page 2021

  3. [3]

    \ Tsvankin, I

    alkhalifah1995velocity APACrefauthors Alkhalifah, T. \ Tsvankin, I. APACrefauthors \ 1995 . Velocity analysis for transversely isotropic media Velocity analysis for transversely isotropic media . Geophysics 60 5 1550--1566

    work page 1995

  4. [4]

    APACrefauthors \ 1981

    berkhout1981wave APACrefauthors Berkhout, A. APACrefauthors \ 1981 . Wave field extrapolation techniques in seismic migration, a tutorial Wave field extrapolation techniques in seismic migration, a tutorial . Geophysics 46 12 1638--1656

    work page 1981

  5. [5]

    , Haghighat, E

    bin2021pinneik APACrefauthors bin Waheed, U. , Haghighat, E. , Alkhalifah, T. , Song, C. \ Hao, Q. APACrefauthors \ 2021 . PINNeik: Eikonal solution using physics-informed neural networks Pinneik: Eikonal solution using physics-informed neural networks . Computers & Geosciences 155 104833

    work page 2021

  6. [6]

    APACrefauthors \ 2010

    carcione2010generalization APACrefauthors Carcione, J M. APACrefauthors \ 2010 . A generalization of the Fourier pseudospectral method A generalization of the fourier pseudospectral method . Geophysics 75 6 A53--A56

    work page 2010

  7. [7]

    , Taufik, M H

    cheng2025seismic APACrefauthors Cheng, S. , Taufik, M H. \ Alkhalifah, T. APACrefauthors \ 2025 . Seismic wavefield solutions via physics-guided generative neural operator Seismic wavefield solutions via physics-guided generative neural operator . arXiv preprint arXiv:2503.06488

    arXiv 2025

  8. [8]

    , Di Cola, V S

    cuomo2022scientific APACrefauthors Cuomo, S. , Di Cola, V S. , Giampaolo, F. , Rozza, G. , Raissi, M. \ Piccialli, F. APACrefauthors \ 2022 . Scientific machine learning through physics--informed neural networks: Where we are and what’s next Scientific machine learning through physics--informed neural networks: Where we are and what’s next . Journal of Sc...

    work page 2022

Show all 39 references
  1. [9]

    , Feng, S

    deng2022openfwi APACrefauthors Deng, C. , Feng, S. , Wang, H. , Zhang, X. , Jin, P. , Feng, Y. Lin, Y. APACrefauthors \ 2022 . OpenFWI: Large-scale multi-structural benchmark datasets for full waveform inversion Openfwi: Large-scale multi-structural benchmark datasets for full...

    2022

  2. [10]

    \ P s enc \' k, I

    gajewski1987computation APACrefauthors Gajewski, D. \ P s enc \' k, I. APACrefauthors \ 1987 . Computation of high-frequency seismic wavefields in 3-D laterally inhomogeneous anisotropic media Computation of high-frequency seismic wavefields in 3-d laterally inhomogeneous anis...

    1987

  3. [11]

    , Duchkov, A

    grubas2023neural APACrefauthors Grubas, S. , Duchkov, A. \ Loginov, G. APACrefauthors \ 2023 . Neural Eikonal solver: Improving accuracy of physics-informed neural networks for solving eikonal equation in case of caustics Neural eikonal solver: Improving accuracy of physics-in...

    2023

  4. [12]

    \ Alkhalifah, T

    huang2022pinnup APACrefauthors Huang, X. \ Alkhalifah, T. APACrefauthors \ 2022 . PINNup: Robust neural network wavefield solutions using frequency upscaling and neuron splitting Pinnup: Robust neural network wavefield solutions using frequency upscaling and neuron splitting ....

    2022

  5. [13]

    \ Alkhalifah, T

    huang2024learned APACrefauthors Huang, X. \ Alkhalifah, T. APACrefauthors \ 2024 . Learned Frequency-Domain Scattered Wavefield Solutions Using Neural Operator Learned frequency-domain scattered wavefield solutions using neural operator . 85th EAGE Annual Conference & Exhibiti...

    2024

  6. [14]

    \ Alkhalifah, T

    huang2025learned APACrefauthors Huang, X. \ Alkhalifah, T. APACrefauthors \ 2025 . Learned frequency-domain scattered wavefield solutions using neural operators Learned frequency-domain scattered wavefield solutions using neural operators . Geophysical Journal International 24...

    2025

  7. [15]

    , Kevrekidis, I G

    karniadakis2021physics APACrefauthors Karniadakis, G E. , Kevrekidis, I G. , Lu, L. , Perdikaris, P. , Wang, S. \ Yang, L. APACrefauthors \ 2021 . Physics-informed machine learning Physics-informed machine learning . Nature Reviews Physics 3 6 422--440

    2021

  8. [16]

    , Kovachki, N

    li2020fourier APACrefauthors Li, Z. , Kovachki, N. , Azizzadenesheli, K. , Liu, B. , Bhattacharya, K. , Stuart, A. \ Anandkumar, A. APACrefauthors \ 2020 1 . Fourier neural operator for parametric partial differential equations Fourier neural operator for parametric partial di...

    2020 arXiv

  9. [17]

    , Kovachki, N

    li2020neural APACrefauthors Li, Z. , Kovachki, N. , Azizzadenesheli, K. , Liu, B. , Bhattacharya, K. , Stuart, A. \ Anandkumar, A. APACrefauthors \ 2020 2 . Neural operator: Graph kernel network for partial differential equations Neural operator: Graph kernel network for parti...

    2020 arXiv

  10. [18]

    , Zheng, H

    li2024physics APACrefauthors Li, Z. , Zheng, H. , Kovachki, N. , Jin, D. , Chen, H. , Liu, B. Anandkumar, A. APACrefauthors \ 2024 . Physics-informed neural operator for learning partial differential equations Physics-informed neural operator for learning partial differential ...

    2024

  11. [19]

    , Jin, P

    lu2021learning APACrefauthors Lu, L. , Jin, P. , Pang, G. , Zhang, Z. \ Karniadakis, G E. APACrefauthors \ 2021 . Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators Learning nonlinear operators via deeponet based on the universa...

    2021

  12. [20]

    APACrefauthors \ 2024

    okazaki2024scientific APACrefauthors Okazaki, T. APACrefauthors \ 2024 . Scientific Machine Learning Seismology Scientific machine learning seismology . arXiv preprint arXiv:2409.18397

    2024 arXiv

  13. [21]

    , Subramanian, S

    pathak2022fourcastnet APACrefauthors Pathak, J. , Subramanian, S. , Harrington, P. , Raja, S. , Chattopadhyay, A. , Mardani, M. others APACrefauthors \ 2022 . Fourcastnet: A global data-driven high-resolution weather model using adaptive fourier neural operators Fourcastnet: A...

    2022 arXiv

  14. [22]

    , Perdikaris, P

    raissi2019physics APACrefauthors Raissi, M. , Perdikaris, P. \ Karniadakis, G E. APACrefauthors \ 2019 . Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations Physics-informed neu...

    2019

  15. [23]

    , Huber, C

    rasht2022physics APACrefauthors Rasht-Behesht, M. , Huber, C. , Shukla, K. \ Karniadakis, G E. APACrefauthors \ 2022 . Physics-informed neural networks (PINNs) for wave propagation and full waveform inversions Physics-informed neural networks (pinns) for wave propagation and f...

    2022

  16. [24]

    , Al Majed, H

    rosofsky2023applications APACrefauthors Rosofsky, S G. , Al Majed, H. \ Huerta, E. APACrefauthors \ 2023 . Applications of physics informed neural operators Applications of physics informed neural operators . Machine Learning: Science and Technology 4 2 025022

    2023

  17. [25]

    ryu2024operator APACrefauthors Ryu, J J. , Xu, X. , Erol, H. , Bu, Y. , Zheng, L. \ Wornell, G W. APACrefauthors \ 2024 . Operator SVD with neural networks via nested low-rank approximation Operator svd with neural networks via nested low-rank approximation . arXiv preprint ar...

    2024 arXiv

  18. [26]

    , Alkhalifah, T

    song2021solving APACrefauthors Song, C. , Alkhalifah, T. \ Waheed, U B. APACrefauthors \ 2021 . Solving the frequency-domain acoustic VTI wave equation using physics-informed neural networks Solving the frequency-domain acoustic vti wave equation using physics-informed neural ...

    2021

  19. [27]

    , Alkhalifah, T

    song2022versatile APACrefauthors Song, C. , Alkhalifah, T. \ Waheed, U B. APACrefauthors \ 2022 . A versatile framework to solve the Helmholtz equation using physics-informed neural networks A versatile framework to solve the helmholtz equation using physics-informed neural ne...

    2022

  20. [28]

    , Zhao, T

    song2024seismic APACrefauthors Song, C. , Zhao, T. , Bin Waheed, U. , Liu, C. \ Tian, Y. APACrefauthors \ 2024 . Seismic traveltime simulation for variable velocity models using physics-informed Fourier neural operator Seismic traveltime simulation for variable velocity models...

    2024

  21. [29]

    \ Alkhalifah, T

    taufik2025wavenumber APACrefauthors Taufik, M H. \ Alkhalifah, T. APACrefauthors \ 2025 . Wavenumber-aware diffusion sampling to regularize multiparameter elastic full waveform inversion Wavenumber-aware diffusion sampling to regularize multiparameter elastic full waveform inv...

    2025

  22. [30]

    , Waheed, U b

    taufik2023neural APACrefauthors Taufik, M H. , Waheed, U b. \ Alkhalifah, T A. APACrefauthors \ 2023 . A neural network based global traveltime function (GlobeNN) A neural network based global traveltime function (globenn) . Scientific Reports 13 1 7179

    2023

  23. [31]

    , Calandra, H

    virieux2011review APACrefauthors Virieux, J. , Calandra, H. \ Plessix, R \'E . APACrefauthors \ 2011 . A review of the spectral, pseudo-spectral, finite-difference and finite-element modelling techniques for geophysical imaging A review of the spectral, pseudo-spectral, finite...

    2011

  24. [32]

    , Zhang, H

    wang2024transfer APACrefauthors Wang, Y. , Zhang, H. , Lai, C. \ Hu, X. APACrefauthors \ 2024 . Transfer learning Fourier neural operator for solving parametric frequency-domain wave equations Transfer learning fourier neural operator for solving parametric frequency-domain wa...

    2024

  25. [33]

    wen2023real APACrefauthors Wen, G. , Li, Z. , Long, Q. , Azizzadenesheli, K. , Anandkumar, A. \ Benson, S M. APACrefauthors \ 2023 . Real-time high-resolution CO 2 geological storage prediction using nested Fourier neural operators Real-time high-resolution co 2 geological sto...

    2023

  26. [34]

    \ Alkhalifah, T

    wu2018efficient APACrefauthors Wu, Z. \ Alkhalifah, T. APACrefauthors \ 2018 1 . An efficient Helmholtz solver for acoustic transversely isotropic media An efficient helmholtz solver for acoustic transversely isotropic media . Geophysics 83 2 C75--C83

    2018

  27. [35]

    \ Alkhalifah, T

    wu2018highly APACrefauthors Wu, Z. \ Alkhalifah, T. APACrefauthors \ 2018 2 . A highly accurate finite-difference method with minimum dispersion error for solving the Helmholtz equation A highly accurate finite-difference method with minimum dispersion error for solving the he...

    2018

  28. [36]

    , Wang, Y

    xue2025multi APACrefauthors Xue, Z. , Wang, Y. , Wu, X. \ Ma, J. APACrefauthors \ 2025 . Multi-geophysical information neural network for seismic tomography Multi-geophysical information neural network for seismic tomography . Geophysics 90 3 R89--R100

    2025

  29. [37]

    , Meng, H

    yang2024frequency APACrefauthors Yang, B. , Meng, H. , Gu, N. , Liu, X. , Chen, X. \ Ben-Zion, Y. APACrefauthors \ 2024 . A frequency domain methodology for quantitative evaluation of diffuse wavefield with applications to seismic imaging A frequency domain methodology for qua...

    2024

  30. [38]

    , Gao, A F

    yang2023rapid APACrefauthors Yang, Y. , Gao, A F. , Azizzadenesheli, K. , Clayton, R W. \ Ross, Z E. APACrefauthors \ 2023 . Rapid seismic waveform modeling and inversion with neural operators Rapid seismic waveform modeling and inversion with neural operators . IEEE Transacti...

    2023

  31. [39]

    , Azizzadenesheli, K

    zou2024deep APACrefauthors Zou, C. , Azizzadenesheli, K. , Ross, Z E. \ Clayton, R W. APACrefauthors \ 2024 . Deep neural Helmholtz operators for 3-D elastic wave propagation and inversion Deep neural helmholtz operators for 3-d elastic wave propagation and inversion . Geophys...

    2024

Pith tools

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