Pith. sign in

REVIEW 4 major objections 4 minor 1 cited by

Deep Reparameterization for Full Waveform Inversion: Architecture Benchmarking, Robust Inversion, and Multiphysics Extension

T0 review · 4 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A CNN that represents the velocity model improves full waveform inversion accuracy and robustness.

desk verdict A useful benchmark study of deep reparameterization for FWI, but warm-start initialization and test-set hyperparameter selection mean the practical claims outrun the evidence. read the letter →

arxiv 2504.17375 v2 pith:G7POULA5 submitted 2025-04-24 physics.geo-ph

Feng Liu , Yaxing Li , Rui Su , Jianping Huang , Lei Bai This is my paper · ORCID
classification physics.geo-ph
keywords deepreparameterizationfullwaveforminversionnetworkarchitecturesearchspectralbiassparseacquisitionmultiparametercrosstalkmitigationseismicimaging
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 argues that representing the subsurface velocity model as the output of a deep neural network, rather than optimizing pixel values directly, makes full waveform inversion more accurate, stable, and tolerant of noisy or sparse data. It systematically benchmarks three network families—U-Net, CNN, and MLP—combined with two ways of injecting a starting model, and finds that a simple CNN with a pretraining-based warm-up works best across three benchmark models. The same reparameterization, in a backbone–branch form, is extended to joint inversion of P-wave velocity, S-wave velocity, and density, and is reported to suppress crosstalk artifacts. The paper also proposes a mechanism: the network's spectral bias imposes a progressive low-to-high frequency learning schedule that mimics multiscale FWI. If these claims hold, the practical consequences are design rules for network-based FWI and reduced data-acquisition costs.

What carries the argument

The load-bearing object is the deep reparameterization map $m = \mathcal{N}(\theta|I_r)$, where a generative network's weights are optimized by the FWI objective. The specific variant that carries the result is CNN-vp: a shallow CNN that takes a fixed random array, passes it through a fully connected layer and convolutional layers, and is warmed up by pretraining to output the Gaussian-smoothed initial model before inversion. In the multiparameter extension, the machinery is a backbone-branch CNN in which a shared backbone extracts common structure and separate branches, each followed by normalization to its physical scale, produce vp, vs, and rho. The paper attributes the method's behaviour to spectral bias—the network learns low-frequency components first, enforcing an implicit progressive multi-scale regularization.

What would settle it

Run CNN-vp and conventional FWI on a synthetic model (e.g., a salt-inclusion model) where the initial model is obtained by smoothing a deliberately wrong large-scale structure, such as a shifted salt body, rather than the true model. If, under the paper's noise and sparsity settings, CNN-vp no longer beats conventional FWI in MAPE or SSIM—or if several CNN architectures with the pretraining warm-up collapse—the claim that reparameterization robustly improves FWI would be refuted. A second check targets the mechanism: on a different model, measure the early-iteration high-frequency energy ratio of conventional FWI; if it does not show the reported abrupt surge, the spectral-bias account needs revision.

Watch

Extended reading notes

Core claim

On its own terms, the central claim is that deep reparameterization—replacing the model $m$ by a generator network output $m = \mathcal{N}(\theta|I_r)$ and optimizing the network weights $\theta$—materially improves FWI, and that the best configuration is a multi-layer CNN pretrained to reproduce the initial smoothed velocity model before physics-driven inversion. The paper reports that this CNN-vp configuration beats traditional FWI, U-Net, and MLP variants on the Marmousi2, Overthrust, and Foothill models in MAPE, SSIM, and SNR, holds up under Gaussian noise up to $6\sigma_0$, and retains an advantage at extreme sparsity (e.g., 2 sources or 20 receivers). For multiparameter FWI, a shared-backbone plus branch CNN with per-parameter normalization and denormalization is claimed to nearly eliminate crosstalk on an anomaly model and substantially improve vp/vs/rho recovery on Marmousi2. The proposed explanation is spectral: conventional FWI injects high-frequency energy abruptly, while the network first dominates low frequencies and expands bandwidth gradually, reducing cycle-skipping.

Load-bearing premise

The benchmarks start every inversion from an initial model obtained by Gaussian-smoothing the true model, and the pretraining step fits the network to that smoothed reference, so the demonstrated superiority over conventional FWI is only shown in a warm-start regime where the large-scale structure of the answer is already known.

Editorial extensions

If this is right

  • A simple CNN with pretraining-based warm-up can be adopted as a plug-in reparameterization module in existing FWI workflows, with architecture guidance that shallower CNNs outperform U-Net and MLP for velocity reconstruction.
  • Pretraining-based initial-model embedding is consistently better than direct perturbation superposition across all tested architectures, so new DR-FWI designs should prefer the warm-up strategy.
  • DR-FWI retains accuracy at acquisition levels where conventional FWI degrades (e.g., 10 sources by 20 receivers), implying reduced field acquisition requirements for similar image quality.
  • In multiparameter elastic inversion, the backbone-branch structure lets vp, vs, and rho be inverted jointly with less crosstalk, without explicit Hessian-based corrections.
  • The spectral-bias explanation predicts that DR-FWI behaves like an adaptive multiscale method, which could reduce the need for manual frequency-continuation strategies.

Reading between the lines

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

  • If the spectral-bias mechanism is the real driver, then network architectures with stronger low-frequency bias should further improve robustness; this can be tested by comparing different activation functions, depths, or Fourier-feature embeddings within the same benchmark.
  • The pretraining warm-up effectively injects the entire initial model as a prior, so the method's edge over conventional FWI may shrink when the initial model is poor; a direct experiment would be to replace the Gaussian-smoothed true model with a biased or outdated velocity model and measure the performance gap.
  • The backbone-branch success suggests a route to multiphysics joint inversion (e.g., elastic plus electromagnetic or gravity), where parameter-specific branches could be governed by different forward operators while the backbone shares structure.
  • Because the network input is fixed and random, the method yields a deterministic map per initialization; ensembling over multiple random inputs could provide cheap uncertainty estimates for the inverted models.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The manuscript proposes DR-FWI, a deep reparameterization framework for full waveform inversion in which subsurface parameters (e.g., vp, vs, rho) are represented by the output of a neural network and the network weights are optimized against the seismic data misfit. It compares three network architectures (CNN, MLP, U-Net) and two reference-embedding strategies (pretraining-based vs. perturbation-based), and reports that a CNN with pretraining-based initialization (CNN-vp) outperforms traditional FWI and the other architecture/strategy combinations on the Marmousi2, Overthrust, and Foothill models. The paper also presents robustness experiments under noise and sparse acquisition, a spectral-bias analysis of the inversion dynamics, and a backbone-branch architecture for multiparameter joint inversion. The authors conclude that DR-FWI provides practical design guidance for network-based FWI, reduces acquisition requirements, and mitigates multiparameter crosstalk.

Significance. If the claims hold, this work provides a systematic and useful comparison of reparameterization architectures and embedding strategies, and it demonstrates a plausible extension to multiparameter FWI. The paper is thorough in its ablation scope, uses the open-source ADFWI toolbox, and reports a variety of quantitative metrics (MAPE, SSIM, SNR). The spectral-bias discussion is a valuable interpretive addition. However, the benchmark methodology has three important weaknesses: the best network configuration is selected on the test model itself, the traditional FWI baseline is unregularized, and the single-parameter experiments are performed only in a warm-start regime where the initial model is a smoothed version of the true model. These issues make the reported quantitative advantages optimistic and limit the strength of the practical design guidance that can be drawn from the current experiments.

major comments (4)
  1. [Section III-B] The architecture comparison is compromised by selecting the best network configuration on the test model. The text states, 'we perform a network architecture search for each architecture and select the best-performing configuration for comparison,' and the same Marmousi2, Overthrust, and Foothill models are then used to report the final results in Table II and Figs. 4-5. This is selection on the test set: each reported number is the best among 8 to 16 random or hyperparameter variants, which inflates measured accuracy and makes the ranking between architectures an upper-bound comparison rather than an unbiased benchmark. Please use a validation model or hold-out split for configuration selection, or report the full distributions rather than only the best run.
  2. [Section III-A and Eq. (11)] The single-parameter experiments are performed only in a warm-start regime. For each model, the initial velocity model is obtained by Gaussian smoothing of the true model, and Eq. (11) pretrains the network to reproduce exactly that smoothed reference. Thus the pretraining target and the starting model contain the correct large-scale structure of the answer, and all reported improvements over traditional FWI are demonstrated only from such warm starts. The paper provides no experiment with an initial model containing low-wavenumber errors (e.g., incorrect velocity trends or misplaced structures), which is the situation encountered in field practice. To support the claimed practical guidance and real-world applicability, please add experiments with erroneous initial models (for example, from travel-time tomography or strongly perturbed smooth models) and show that the inversion can correct, or at least copes with, an incorrect reference.
  3. [Section III-A and Table II] The traditional FWI baseline uses no regularization, multiscale continuation, or similar safeguards, even though the paper attributes the advantage of DR-FWI largely to implicit regularization. This makes it unclear how much of the reported improvement comes from the network reparameterization itself and how much from the absence of standard regularization in the baseline. Please add a conventional regularized or multiscale baseline (e.g., total-variation regularization or frequency continuation) to the comparisons in Table II and the robustness curves in Figs. 6-7, or justify why the unregularized baseline is the appropriate reference.
  4. [Table II and Figs. 6-7] All quantitative results appear to be single runs per configuration. The shaded regions in Figs. 5-7 show variability across network hyperparameters, not across random initializations or optimizer stochasticity. Since the pretraining and the subsequent inversion are stochastic, the reported differences (e.g., CNN-vp versus MLP-vp, or DR-FWI versus traditional FWI) may not be statistically stable. Please report the mean and standard deviation over multiple independent random seeds for the headline comparisons, and state whether the differences are significant.
minor comments (4)
  1. [Section II-B, Eq. (5)] The update rule mk+1 = mk + alpha_k * partial L / partial mk has the wrong sign for gradient descent; as written it performs ascent on the objective unless partial L / partial mk is defined as the negative gradient. Please correct the equation or clarify the sign convention.
  2. [Section IV-A, Eq. (17)] The high-frequency ratio threshold rc = 2.375 is introduced without derivation or sensitivity analysis. Please define the frequency units and report how the qualitative conclusions of Fig. 10 depend on the choice of rc.
  3. [Introduction and Limitations] The Introduction criticizes the absence of open-source implementations, but no code is released for the proposed DR-FWI framework itself; the paper only cites the ADFWI toolbox. Please provide a public repository with the network architectures and training/inversion scripts.
  4. [Throughout] There are several minor grammatical errors and typos, such as 'supported by the the Shanghai' in the footnote and 'this results highlight' in Section III-D. A careful proofread is recommended.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the final inversion is driven by seismic data misfit, and no equation reduces the predictions to the pretraining reference or to a self-citation.

full rationale

The paper's central claims are empirical comparisons obtained by optimizing the FWI data-misfit objective after reparameterizing the model as m = N(theta|Ir) (Eq. 6) and solving Eq. 7, i.e., minimizing D(dobs, dcal(N(theta|Ir))) with respect to the network weights. The pretraining step (Eq. 11) only initializes the network to reproduce the Gaussian-smoothed starting model m0; it does not enter the inversion loss as a fixed target, and the final result is therefore not equal to m0 by construction. The warm-start protocol described in Section III-A ('the corresponding initial velocity model is obtained by applying a Gaussian smoothing operation') is a limitation on external validity, not a circular reduction: traditional FWI starts from the same smoothed model, and the CNN-vp comparisons are conducted under identical conditions, so the relative gains are not baked into the pretraining target. The spectral-bias discussion in Section IV-A uses a hand-picked threshold rc=2.375 in the HF-Ratio diagnostic (Eq. 17), but this diagnostic merely describes the optimization trajectory and is not used in the objective; a threshold choice is not circularity. The self-citations to Liu et al. (2025) [24] and the ADFWI toolbox [34] are implementation and software references rather than load-bearing mathematical assumptions; no uniqueness theorem or ansatz is imported from the authors' prior work to force the architecture conclusions. Because the quantitative results come from external benchmark models (Marmousi2, Overthrust, Foothill) and the derivations do not reduce to the inputs, no circular step can be exhibited.

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

The method contributes a benchmark and architecture variant; the free parameters above capture the design choices that the reported numbers depend on but are not fixed by first principles. No new physical entities, forces, or conserved quantities are introduced.

free parameters (4)
  • HF-Ratio threshold rc = 2.375
    Hand-set frequency boundary in Eq. 17 used to define the spectral evolution claim; no sensitivity analysis or physical justification is given.
  • Best network configuration per architecture = not reported
    Architecture search over layers and channels selects the winning CNN/MLP/U-Net variant per test model (Section III-B), but the exact configuration is not listed, making the reported numbers conditional on an unspecified choice.
  • Learning rate and StepLR schedule = not reported
    Section III-A states the learning rate is tuned and StepLR is used, but no values are given; these affect convergence of all methods.
  • Gaussian smoothing window size for initial model = 240 m x 240 m to 300 m x 300 m
    The initial reference model is the true model blurred with this window (Table I); all results depend on this warm-start condition.
assumptions (4)
  • standard math Acoustic and elastic wave equations (Eqs. 1-2) model seismic wave propagation.
    Standard physics assumed throughout; forward modeling uses ADFWI.
  • domain assumption Deep Image Prior assumption that network architecture alone provides a usable structural prior (Ulyanov et al. 2018).
    Invoked in Section II-B as the basis for reparameterization.
  • domain assumption Global-correlation misfit with Adam optimization converges to meaningful solutions in 300 iterations.
    Experimental design choice; no proof of convergence, but standard practice in FWI studies.
  • domain assumption Synthetic benchmark models (Marmousi2, Overthrust, Foothill) are representative of real subsurface complexity.
    Used to generalize conclusions to field applications (Section IV-C).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deep Reparameterization for Full Waveform Inversion: Architecture Benchmarking, Robust Inversion, and Multiphysics Extension." pith.science (2026). https://pith.science/paper/G7POULA5

@misc{pith2026250417375,
  author       = {Pith},
  title        = {Pith review of: Deep Reparameterization for Full Waveform Inversion: Architecture Benchmarking, Robust Inversion, and Multiphysics Extension},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G7POULA5}},
  note         = {Machine review of arXiv:2504.17375}
}
read the original abstract

Full waveform inversion (FWI) is a high-resolution subsurface imaging technique, but its effectiveness is limited by challenges such as noise contamination, sparse acquisition, and artifacts from multiparameter coupling. To address these limitations, this study develops a deep reparameterized FWI (DR-FWI) framework, in which subsurface parameters are represented by a deep neural network. Instead of directly optimizing the parameters, DR-FWI optimizes the network weights to reconstruct them, thereby embedding structural priors and facilitating optimization. To provide benchmark guidelines for the design of DR-FWI, we conduct a comparative analysis of three representative architectures (U-Net, CNN, MLP) combined with two initial model embedding strategies: one pretraining the network to generate predefined initial models (pretraining-based), while the other directly adds network outputs to the initial models. Extensive ablation experiments show that combining CNN with pretraining-based initialization significantly enhances inversion accuracy, offering valuable insights into network design. To further understand the mechanism of DR-FWI, spectral bias analysis reveals that the network first captures low-frequency features and gradually reconstructs high-frequency details, enabling an adaptive multi-scale inversion strategy. Notably, the robustness of DR-FWI is validated under various noise levels and sparse acquisition scenarios, where its strong performance with limited shots and receivers demonstrates reduced reliance on dense observational data. Additionally, a backbone-branch structure is proposed to extend DR-FWI to multiparameter inversion, and its efficacy in mitigating cross-parameter interference is validated on a synthetic anomaly model and the Marmousi2 model. These results suggest a promising direction for joint inversion involving multiple parameters or multiphysics.

Figures

Figures reproduced from arXiv: 2504.17375 by the authors.

Figure 1
Figure 1. Comparison of data-driven methods (e.g., generative model-based, end-to-end supervised learning) and physics-driven methods (e.g., traditional FWI, [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Workflow of deep reparameterization-based full waveform inversion. The left panel illustrates three types of reparameterization neural networks (U-Net, [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Overview of the ”backbone-branch” architecture for multiparameter [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Comparison of inversion results using different deep reparame [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Comparison of inversion performance metrics for different deep [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Robustness evaluation of deep reparameterized FWI under noisy [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Robustness evaluation of deep reparameterized FWI under sparse-data [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Multiparameter inversion results on an anomaly model. (a–c) Ground [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: Multiparameter inversion on Marmousi2 model. (a–c) Ground-truth [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 12
Figure 12. Figure 12: Comparison of MAPE evolution during inversion for (a) [PITH_FULL_IMAGE:figures/full_fig_p011_12.png]
Figure 11
Figure 11. Figure 11: Comparison of inversion results for vp, vs, and ρ obtained from two reparameterization strategies in multiparameter FWI: a single ”backbone￾branch” network (a, c, e) and three independent CNN networks (b, d, f). Subfigures (a, b) correspond to vp, (c, d) to vs, and (e…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. A Memory Efficient Adjoint Method to Enable Billion Parameter Optimization on a Single GPU in Dynamic Problems

    cs.CE 2025-09 conditional novelty 7.0 of 10

    A superposition of forward and adjoint wavefields with a scaling factor lets the authors approximate the optimization sensitivity using O(number of grid points) memory instead of O(steps times grid points).

Reference graph

Works this paper leans on

54 extracted references · 47 canonical work pages · cited by 1 Pith paper

  1. [1]

    The seismic inverse problem as a sequence of before stack mi- grations,

    P. Lailly, “The seismic inverse problem as a sequence of before stack mi- grations,” in Conference on Inverse Scattering, Theory and Application, Society for Industrial and Applied Mathematics, Expanded Abstracts , Philadelphia, 1983, pp. 206–220

    work page 1983

  2. [2]

    Linearized inversion of seismic reflection data,

    A. Tarantola, “Linearized inversion of seismic reflection data,” Geophys- ical Prospecting, vol. 32, no. 6, pp. 998–1015, 1984

    work page 1984

  3. [3]

    An overview of full-waveform inversion in exploration geophysics,

    J. Virieux and S. Operto, “An overview of full-waveform inversion in exploration geophysics,” Geophysics, vol. 74, no. 6, pp. WCC1–WCC26, 2009

    2009

  4. [4]

    Seismic wavefield imaging of earth’s interior across scales,

    J. Tromp, “Seismic wavefield imaging of earth’s interior across scales,” Nature Reviews Earth and Environment , vol. 1, no. 1, pp. 40–53, 2019

    work page 2019

  5. [5]

    Regularized full- waveform inversion with automated salt flooding,

    M. Kalita, V . Kazei, Y . Choi, and T. Alkhalifah, “Regularized full- waveform inversion with automated salt flooding,” Geophysics, vol. 84, no. 4, pp. R569–R582, 2019

    work page 2019

  6. [6]

    Seismic imaging of complex structures by 2d elastic frequency-domain full-waveform inversion,

    R. Brossier, S. Operto, and V . Jean, “Seismic imaging of complex structures by 2d elastic frequency-domain full-waveform inversion,” Geophysics, vol. 74, no. 6, 2009

    work page 2009

  7. [7]

    Impact of high-resolution fwi in the western black sea: Revealing overburden and reservoir complexity,

    P. Routh, R. Neelamani, R. Lu, S. Lazaratos, H. Braaksma, S. Hughes, R. Saltzer, J. Stewart, K. Naidu, H. Averill, V . Gottumukkula, P. Homonko, J. Reilly, and D. Leslie, “Impact of high-resolution fwi in the western black sea: Revealing overburden and reservoir complexity,” The Leading Edge , vol. 36, no. 1, pp. 60–66, 2017

    work page 2017

  8. [8]

    Advanced seismic processing/imaging tech- niques and their potential for geothermal exploration,

    C. Schmelzbach, S. Greenhalgh, F. Reiser, J.-F. Girard, F. Bretaudeau, L. Capar, and A. Bitri, “Advanced seismic processing/imaging tech- niques and their potential for geothermal exploration,” Interpretation, vol. 4, no. 4, pp. SR1–SR18, 2016

    work page 2016

Show all 54 references
  1. [9]

    Two-dimensional nonlinear inversion of seismic waveforms: Numerical results,

    O. Gauthier, J. Virieux, and A. Tarantola, “Two-dimensional nonlinear inversion of seismic waveforms: Numerical results,”Geophysics, vol. 51, no. 7, pp. 1387–1403, 1986

    1986

  2. [10]

    Wave-equation traveltime inversion,

    Y . Luo and G. T. Schuster, “Wave-equation traveltime inversion,” Geo- physics, vol. 56, no. 5, pp. 645–653, 1991

    1991

  3. [11]

    Full waveform inversion and the truncated newton method: Quantitative imaging of complex subsurface structures,

    L. M ´etivier, F. Bretaudeau, R. Brossier, S. Operto, and J. Virieux, “Full waveform inversion and the truncated newton method: Quantitative imaging of complex subsurface structures,” Geophysical Prospecting , vol. 62, no. 6, pp. 1353–1375, 2014

    2014

  4. [12]

    Tarantola, Inverse Problem Theory and Methods for Model Parameter Estimation

    A. Tarantola, Inverse Problem Theory and Methods for Model Parameter Estimation. Society for Industrial and Applied Mathematics, 2005

    2005

  5. [13]

    A guided tour of multiparameter full-waveform inversion with multicomponent data: From theory to practice,

    S. Operto, Y . Gholami, V . Prieux, A. Ribodetti, R. Brossier, L. Metivier, and J. Virieux, “A guided tour of multiparameter full-waveform inversion with multicomponent data: From theory to practice,” The Leading Edge , vol. 32, no. 9, pp. 1040–1054, 2013

    2013

  6. [14]

    A multi-parameter full waveform inversion strategy for acoustic media with variable density,

    J. Yang, Y . Liu, and L. Dong, “A multi-parameter full waveform inversion strategy for acoustic media with variable density,” Chinese Journal of Geophysics (in Chinese) , vol. 57, no. 2, pp. 628–643, 2014

    2014

  7. [15]

    Deep learning for geophysics: Current and future trends,

    S. Yu and J. Ma, “Deep learning for geophysics: Current and future trends,” Reviews of Geophysics, vol. 59, no. 3, p. e2021RG000742, 2021

    2021

  8. [16]

    Physics-guided data-driven seismic inversion: Recent progress and future opportunities in full-waveform inversion,

    Y . Lin, J. Theiler, and B. Wohlberg, “Physics-guided data-driven seismic inversion: Recent progress and future opportunities in full-waveform inversion,” IEEE Signal Processing Magazine , vol. 40, no. 1, pp. 115– 133, 2023

    2023

  9. [17]

    Inversionnet: An efficient and accurate data-driven full waveform inversion,

    Y . Wu and Y . Lin, “Inversionnet: An efficient and accurate data-driven full waveform inversion,”IEEE Transactions on Computational Imaging, vol. 6, pp. 419–433, 2020

    2020

  10. [18]

    Data-driven seismic waveform inversion: A study on the robustness and generalization,

    Z. Zhang and Y . Lin, “Data-driven seismic waveform inversion: A study on the robustness and generalization,” IEEE Transactions on Geoscience and Remote Sensing , vol. 58, no. 10, pp. 6900–6913, 2020

    2020

  11. [19]

    Physics-informed neural networks (pinns) for wave propagation and full waveform inversions,

    M. Rasht-Behesht, C. Huber, K. Shukla, and G. E. Karniadakis, “Physics-informed neural networks (pinns) for wave propagation and full waveform inversions,” Journal of Geophysical Research: Solid Earth , vol. 127, no. 5, 2022

    2022

  12. [20]

    A physics-based neural- network way to perform seismic full waveform inversion,

    Y . Ren, X. Xu, S. Yang, L. Nie, and Y . Chen, “A physics-based neural- network way to perform seismic full waveform inversion,” IEEE Access, vol. 8, pp. 112 266–112 277, 2020

    2020

  13. [21]

    Reparameterized full-waveform inversion using deep neural networks,

    Q. He and Y . Wang, “Reparameterized full-waveform inversion using deep neural networks,” Geophysics, vol. 86, no. 1, pp. V1–V13, 2021

    2021

  14. [22]

    Integrating deep neural networks with full-waveform inversion: Reparameterization, regularization, and uncertainty quantification,

    W. Zhu, K. Xu, E. Darve, B. Biondi, and G. C. Beroza, “Integrating deep neural networks with full-waveform inversion: Reparameterization, regularization, and uncertainty quantification,” Geophysics, vol. 87, no. 1, pp. R93–R109, 2021

    2021

  15. [23]

    Implicit seismic full wave- form inversion with deep neural representation,

    J. Sun, K. Innanen, T. Zhang, and D. Trad, “Implicit seismic full wave- form inversion with deep neural representation,” Journal of Geophysical Research: Solid Earth , vol. 128, no. 3, p. e2022JB025964, 2023

    2023

  16. [24]

    Automatic differentiation-based full waveform inversion with flexible workflows,

    F. Liu, H. Li, G. Zou, and J. Li, “Automatic differentiation-based full waveform inversion with flexible workflows,” Journal of Geophysi- cal Research: Machine Learning and Computation , vol. 2, no. 1, p. e2024JH000542, 2025

    2025

  17. [25]

    Parametric convolutional neural network- domain full-waveform inversion,

    Y . Wu and G. A. McMechan, “Parametric convolutional neural network- domain full-waveform inversion,” Geophysics, vol. 84, no. 6, pp. R881– R896, 2019

    2019

  18. [26]

    Physics-guided deep learning for seismic inversion with hybrid training and uncertainty analysis,

    J. Sun, K. A. Innanen, and C. Huang, “Physics-guided deep learning for seismic inversion with hybrid training and uncertainty analysis,” Geophysics, vol. 86, no. 3, pp. R303–R317, 2021

    2021

  19. [27]

    A prior regularized full wave- form inversion using generative diffusion models,

    F. Wang, X. Huang, and T. A. Alkhalifah, “A prior regularized full wave- form inversion using generative diffusion models,” IEEE Transactions on Geoscience and Remote Sensing , vol. 61, pp. 1–11, 2023

    2023

  20. [28]

    Learned regularizations for multi-parameter elastic full waveform inversion using diffusion models,

    M. H. Taufik, F. Wang, and T. Alkhalifah, “Learned regularizations for multi-parameter elastic full waveform inversion using diffusion models,” Journal of Geophysical Research: Machine Learning and Computation , vol. 1, no. 1, p. e2024JH000125, 2024

    2024

  21. [29]

    Elastic-adjointnet: A physics-guided deep au- toencoder to overcome crosstalk effects in multiparameter full-waveform inversion,

    A. Dhara and M. Sen, “Elastic-adjointnet: A physics-guided deep au- toencoder to overcome crosstalk effects in multiparameter full-waveform inversion,” in Second International Meeting for Applied Geoscience & Energy. Houston, Texas: Society of Exploration Geophysicists and Ame...

    2022

  22. [30]

    Accuracy of finite- difference modeling of the acoustic wave equation,

    R. M. Alford, K. R. Kelly, and D. M. Boore, “Accuracy of finite- difference modeling of the acoustic wave equation,” Geophysics, vol. 39, no. 6, pp. 834–842, 1974

    1974

  23. [31]

    G. T. Schuster, Seismic Inversion. Society of Exploration Geophysicists, 2017

    2017

  24. [32]

    Fourth-order finite-difference p-sv seismograms,

    A. R. Levander, “Fourth-order finite-difference p-sv seismograms,” Geo- physics, vol. 53, no. 11, pp. 1425–1436, 1988

    1988

  25. [33]

    P-SV wave propagation in heterogeneous media: Velocity- stress finite-difference method,

    J. Virieux, “ P-SV wave propagation in heterogeneous media: Velocity- stress finite-difference method,” Geophysics, vol. 51, no. 4, pp. 889–901, 1986

    1986

  26. [34]

    Liufeng2317/adfwi: Zenodo,

    F. Liu, “Liufeng2317/adfwi: Zenodo,” Zenodo, 2024

    2024

  27. [35]

    Misfit functions for full waveform inversion based on instantaneous phase and envelope mea- surements: Misfit functions for full waveform inversion,

    E. Bozda ˘g, J. Trampert, and J. Tromp, “Misfit functions for full waveform inversion based on instantaneous phase and envelope mea- surements: Misfit functions for full waveform inversion,” Geophysical Journal International, vol. 185, no. 2, pp. 845–870, 2011

    2011

  28. [36]

    Application of multi-source waveform inversion to marine streamer data using the global correlation norm,

    Y . Choi and T. Alkhalifah, “Application of multi-source waveform inversion to marine streamer data using the global correlation norm,” Geophysical Prospecting, vol. 60, no. 4, pp. 748–758, 2012

    2012

  29. [37]

    The adjoint method in seismol- ogy,

    A. Fichtner, H.-P. Bunge, and H. Igel, “The adjoint method in seismol- ogy,” Physics of the Earth and Planetary Interiors , vol. 157, no. 1-2, pp. 86–104, 2006

    2006

  30. [38]

    Finite-frequency kernels based on adjoint meth- ods,

    Q. Liu and J. Tromp, “Finite-frequency kernels based on adjoint meth- ods,” Bulletin of the Seismological Society of America , vol. 96, no. 6, pp. 2383–2397, 2006

    2006

  31. [39]

    Au- tomatic differentiation in geophysical inverse problems,

    M. Sambridge, P. Rickwood, N. Rawlinson, and S. Sommacal, “Au- tomatic differentiation in geophysical inverse problems,” Geophysical Journal International, vol. 170, no. 1, pp. 1–8, 2007

    2007

  32. [40]

    On the use of neural networks for full waveform inversion,

    L. Herrmann, T. B ¨urchner, F. Dietrich, and S. Kollmannsberger, “On the use of neural networks for full waveform inversion,” Computer Methods in Applied Mechanics and Engineering , vol. 415, p. 116278, 2023

    2023

  33. [41]

    Deep image prior,

    V . Lempitsky, A. Vedaldi, and D. Ulyanov, “Deep image prior,” in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2018, pp. 9446–9454

    2018

  34. [42]

    Full-waveform inversion using a learned regularization,

    P. Sun, F. Yang, H. Liang, and J. Ma, “Full-waveform inversion using a learned regularization,” IEEE Transactions on Geoscience and Remote Sensing, vol. 61, pp. 1–15, 2023

    2023

  35. [43]

    Reference-driven compressed sensing mr image reconstruction using deep convolutional neural networks without pre-training,

    D. Zhao, F. Zhao, and Y . Gan, “Reference-driven compressed sensing mr image reconstruction using deep convolutional neural networks without pre-training,” Sensors, vol. 20, no. 1, p. 308, 2020

    2020

  36. [44]

    Marmousi2: An elastic upgrade for marmousi,

    G. S. Martin, R. Wiley, and K. J. Marfurt, “Marmousi2: An elastic upgrade for marmousi,” The Leading Edge , vol. 25, no. 2, pp. 156–166, 2006

    2006

  37. [45]

    Aminsadeh, 3-D Salt and Overthrust Seismic Models , P

    F. Aminsadeh, 3-D Salt and Overthrust Seismic Models , P. Weimer and T. L. Davis, Eds. American Association of Petroleum Geologists, 1996

    1996

  38. [46]

    Migration from topography: Improving the near-surface image,

    S. H. Gray and K. J. Marfurt, “Migration from topography: Improving the near-surface image,” Canadian Society of Exploration Geophysicists , vol. 31, pp. 18–24, 1995

    1995

  39. [47]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” 2017

    2017

  40. [48]

    Another look at measures of forecast accuracy,

    R. J. Hyndman and A. B. Koehler, “Another look at measures of forecast accuracy,” International Journal of F orecasting, vol. 22, no. 4, pp. 679– 688, 2006

    2006

  41. [49]

    Image quality assessment: From error visibility to structural similarity,

    Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, “Image quality assessment: From error visibility to structural similarity,” IEEE Transactions on Image Processing , vol. 13, no. 4, pp. 600–612, 2004

    2004

  42. [50]

    On the spectral bias of neural networks,

    N. Rahaman, A. Baratin, D. Arpit, F. Draxler, M. Lin, F. A. Hamprecht, Y . Bengio, and A. Courville, “On the spectral bias of neural networks,” 2018

    2018

  43. [51]

    The spectral bias of the deep image prior,

    P. Chakrabarty and S. Maji, “The spectral bias of the deep image prior,” 2019

    2019

  44. [52]

    Full wave- form inversion of sh- and love-wave data in near-surface prospecting,

    E. Dokter, D. K ¨ohn, D. Wilken, D. De Nil, and W. Rabbel, “Full wave- form inversion of sh- and love-wave data in near-surface prospecting,” Geophysical Prospecting, vol. 65, no. S1, pp. 216–236, 2017

    2017

  45. [53]

    Full-waveform inversion with extrapolated low-frequency data,

    Y . E. Li and L. Demanet, “Full-waveform inversion with extrapolated low-frequency data,” Geophysics, vol. 81, no. 6, pp. R339–R348, 2016

    2016

  46. [54]

    Velocity model building by 3d frequency-domain, full-waveform inversion of wide-aperture seismic data,

    H. Ben-Hadj-Ali, S. Operto, and J. Virieux, “Velocity model building by 3d frequency-domain, full-waveform inversion of wide-aperture seismic data,” Geophysics, vol. 73, no. 5, pp. VE101–VE117, 2008

    2008

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.