Pith. sign in

REVIEW 4 major objections 5 minor 48 references

Acoustic Waveform Inversion with Image-to-Image Schr\"odinger Bridges

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A conditional Schrödinger bridge turns a smoothed velocity model into a posterior sample over ground truth using only a few network evaluations.

desk verdict A credible application paper that borrows its theory from cDSB/I2SB and delivers a useful FWI demo, but the headline posterior guarantee only holds under a distortion prior the authors admit is unrealistic. read the letter →

arxiv 2506.15346 v1 pith:TR5VXVY7 submitted 2025-06-18 cs.LG

A.S. Stankevich , I.B. Petrov This is my paper · ORCID
classification cs.LG
keywords fullwaveforminversionSchrödingerbridgeconditionalgenerativemodelingseismicvelocityestimationimage-to-imagetranslationdiffusionmodelsclassifier-freeguidanceOpenFWI
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 tries to establish that the conditional Image-to-Image Schrödinger Bridge (cI2SB) gives acoustic full waveform inversion a principled way to use a smoothed starting velocity model: inference begins at that smoothed model and ends, in finite time, at a sample from the posterior over true velocity models given the observed seismograms. The authors derive cI2SB from the amortized conditional Schrödinger bridge problem, adapt the I2SB training pipeline to conditional simulation, and add a classifier-free-guidance-style weighting that trades sample diversity against fidelity. On the OpenFWI benchmark, the bridge outperforms a reimplemented conditional diffusion model and matches or beats a supervised baseline in perceptual quality while needing only a handful of neural function evaluations. The payoff, if the claim holds, is that diffusion-based inversion no longer requires a long noising chain to a Gaussian prior or a heuristic recipe for injecting the initial model.

What carries the argument

The central object is the conditional Image-to-Image Schrödinger Bridge (cI2SB): a Schrödinger bridge interpolating between the distribution of ground-truth velocity models and the distribution of smoothed velocity models, with the observed seismic data serving as a conditioning variable that does not evolve along the bridge (dy = 0). The load-bearing identity is the analytic posterior q(xt|x0,x1) of the I2SB formulation, which lets the model sample intermediate velocity states directly during training and gives the DDPM-style recursion used at inference. The training procedure (Alg. 2) learns conditional and unconditional drifts in a single network, alternating between prediction with and without the data channel, and the sampling procedure (Alg. 3) blends the two predictions with a guidance scale η. This machinery converts the previously heuristic 'how to include the initial velocity model' into a boundary condition of a stochastic optimal transport problem.

What would settle it

Run the released cI2SB model starting from a smooth velocity model produced by a different mechanism than pdist — for example, migration or traveltime tomography, or a Gaussian blur with kernel size 16–24 or 0–8 — and compare MAE/MSE/SSIM with the in-distribution results in Tables 2–3; a drop to the levels of Table 6's out-of-distribution rows would falsify the claim that the inference is a general posterior sampler rather than a memorization of the training distortion.

Watch

Extended reading notes

Core claim

The central claim is that the cI2SB framework, derived from the amortized conditional Schrödinger bridge problem, provides an acoustic FWI scheme whose inference starts from a smoothed velocity model and terminates at a sample from the posterior over ground-truth velocity models conditioned on observed seismic data, and that on OpenFWI this scheme outperforms a reimplemented conditional diffusion model while reaching sample fidelity superior to a supervised approach with only a few neural function evaluations. The bridge is constructed by fixing the forward reference measure to the acoustic velocity field while the seismic data channel stays constant, and by taking the endpoint prior to be the parametric distortion pdist(c1|c0) applied to ground truth. Because the bridge posterior q(ct|c0,c1) is analytic, training reduces to denoising score matching with paired (c0, c1, dobs) triples, and inference runs the DDPM recursion backward from csmooth. The authors report that the guidance scale η controls the diversity of reconstructed velocity fields, and that objective reweighting (wcond) is necessary to keep the network attending to the seismic data.

Load-bearing premise

The load-bearing premise is that a real-world smoothed velocity model is a sample from the Gaussian-blur-plus-noise distortion family used in training; the paper states this construction has no basis in real-world data, and its Table 6 shows performance degrades when the smoothing kernel leaves the training range.

Editorial extensions

If this is right

  • Diffusion-based acoustic inversion can start from an arbitrary smoothed velocity model and still be guaranteed, in the continuous-time limit, to produce a posterior sample over true velocity models given the observed data.
  • The cost of high perceptual fidelity drops from hundreds of network evaluations (typical for Gaussian-prior diffusion) to roughly 5–50, making the method practical for large 2D surveys.
  • The guidance scale η gives practitioners a single hyperparameter to trade sample diversity against fidelity, with η = 1 collapsing the output to a near-deterministic reconstruction.
  • Because the distortion operator is explicit, the same framework can be retargeted to other 'smooth start' priors — migration images, tomography models — by retraining with a matching pdist.
  • At matched NFEs, the bridge model consistently beats the conditional diffusion baseline on all tested OpenFWI dataset families by a statistically significant margin.

Reading between the lines

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

  • We infer that the same conditional-bridge construction should transfer to other seismic inverse problems (elastic FWI, migration-image conditioning) wherever a paired smooth-to-true training set exists, since the method only needs pjoin and pdist.
  • The paper's Table 6 suggests a practical calibration recipe: before field deployment, measure the actual distortion between available smooth models and true velocity using wells, and retrain the bridge with that empirical pdist; otherwise the endpoint guarantee does not apply.
  • The authors' observation that the network ignores the data channel unless wcond is large hints at a general pitfall: with informative priors, joint conditional/unconditional training is biased toward the unconditional solution, and explicit loss reweighting may be needed in any bridge-based posterior sampler.
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 / 5 minor

Summary. The paper proposes cI2SB, a conditional Image-to-Image Schrödinger Bridge for acoustic full waveform inversion. It adapts the I2SB training objective to take seismic data and a smoothed velocity model as conditional inputs, and uses a classifier-free guidance style weighting to trade off conditional and unconditional denoising. The inference starts from a smoothed velocity model and terminates at a sample from the learned posterior over velocity models conditioned on the seismic data. Experiments on OpenFWI subsets compare the method against InversionNet and a reimplemented conditional SGM, reporting improved SSIM and sample fidelity with a few NFEs. The paper also studies the effect of NFE count and guidance scale.

Significance. If the empirical claims hold, the paper provides a more principled way to incorporate initial velocity models into diffusion-based FWI than current heuristics, and the few-NFE property is practically valuable. The derivation is a straightforward application of the cDSB framework to the I2SB setting, and the authors are transparent about the synthetic nature of the distortion prior. Strengths include a reproducible code repository, a clean statement of the theoretical backbone, and a systematic NFE study. However, the central guarantee is conditional on the training distortion operator, which the authors admit has no real-world basis; this limits the significance unless addressed.

major comments (4)
  1. [Section 5.1.3 and Section 6.4] The distortion operator D^γ_k of Eq. (25) is explicitly described as 'not hav[ing] any basis in real-world data', and Table 6 shows that using different kernel sizes degrades performance. The theoretical guarantee that the inference process ends at a posterior sample (stated in the Abstract and Section 3) holds only when the initial smoothed model is drawn from the same parametric distortion family used in training. For real FWI initial models from migration or tomography, whose errors are structured and spatially correlated, there is no reason the bridge endpoint will be the posterior. The paper should either incorporate a physics-based data-consistency term into the loss or sampling, or explicitly reframe the main claim as conditional on the distortion model.
  2. [Section 5.2 and Tables 2-3] The claim that cI2SB 'outperforms' the conditional SGM 'by statistically significant margin' is not supported by the reported results. No error bars or repeated runs are provided for the deterministic baselines (InversionNet and cSGM); only the stochastic cI2SB results are averaged over 50 seeds, but their variance is not reported. Moreover, in Table 3, for FlatVelA the cSGM reports lower MAE and MSE than cI2SB OT-ODE sampling, so the statement 'in all studied cases' is factually inaccurate. Please provide confidence intervals or a proper statistical comparison, and soften the claim accordingly.
  3. [Tables 4 and 5] The NFE study contains numerical inconsistencies that undermine the quantitative claims. For example, in Table 4 the FlatVelB row at NFE=1 lists MAE=1.63e-2 and MSE=1.71e-2, which is impossible for normalized data; in Table 5 the FlatFaultB row at NFE=50 lists MAE=3.46e-3, an order of magnitude lower than the neighboring rows, suggesting a missing digit. Please correct these values and re-verify all numbers.
  4. [Section 6.2] The claim that cI2SB achieves superior perceptual quality with fewer NFEs is supported only by visual inspection of Fig. 12. To make this claim quantitative, please report a perceptual metric such as FID or LPIPS, or at least report the variance of the standard metrics over the validation set.
minor comments (5)
  1. [Equation (17)] The variance terms σ^2_t and σ^2_t use identical notation, which is ambiguous. Please follow the I2SB convention of σ_t^2 and σ̄_t^2.
  2. [Section 5.1.4] The description of the InversionNet reimplementation as passing a constant timestep value to the network is unclear; InversionNet is not a diffusion model. Please clarify the exact architecture and training procedure used.
  3. [Section 5.1.1 vs. Supplementary Figure 8] The text reports wcond=100, while the caption of Figure 8 refers to wcond=10. Please reconcile this inconsistency.
  4. [Throughout] There are numerous typos, e.g., 'existense' in Section 1 and inconsistent rendering of 'Schrödinger' as 'Schr¨ odinger'.
  5. [Section 5.1.5] For the stochastic cI2SB results averaged over 50 seeds, please report standard deviations or interquartile ranges in addition to the means.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the cI2SB derivation is grounded in external cited theory, the training loss regresses ground truth directly, and evaluation uses external OpenFWI validation splits.

full rationale

The paper does not exhibit circular derivation. The conditional Schrödinger bridge formulation (Eqs. 13-14 and 21-24) is attributed to Shi et al. [24], and the analytic bridge posterior (Eq. 17) and the delta-measure construction in Section 2.5 are attributed to Liu et al. [25]; these are external peer-reviewed results, and the reference list contains no self-citations by Stankevich/Petrov, so no self-citation chain is load-bearing. The training objective (Algorithms 1-2) directly regresses the ground-truth velocity model c0 from the noisy bridge sample and the conditioning seismogram, and the evaluation in Tables 2-3 is performed against OpenFWI validation splits that were not used to fit model parameters; no fitted parameter is renamed as a prediction. The only qualification is an explicit and acknowledged limitation: the distortion operator in Eq. 25 is a synthetic construction ('Note, however, that suggested construction is generic and does not have any basis in real-world data', Section 5.1.3), and Section 6.4 shows that out-of-distribution smoothed models degrade inversion metrics. That is a correctness and generalization risk for real FWI inputs, not a circularity, because the paper openly states the operator's synthetic nature and does not claim the theoretical endpoint guarantee transfers to arbitrary real-world smoothed models. Thus the derivation chain is self-contained with respect to its own stated assumptions.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on (i) SB machinery borrowed from [24] and [25], (ii) a hand-designed degradation operator that defines the bridge prior and is a free-parameter family with no real-world calibration, (iii) joint conditional/unconditional training with reweighting and a convex guidance rule that lacks rigorous theory, and (iv) a fixed-shape seismogram representation. No new physical entities are postulated. The degradation operator is the dominant non-standard input: the paper's own Section 6.4 shows the method is sensitive to it.

free parameters (4)
  • Distortion operator hyperparameters (k, gamma) = k in [8,16]; gamma in [0,0.2]
    Define the bridge prior pdist(c1|c0) in Eq. (25), Section 5.1.3; hand-chosen, no real-world calibration, and the paper admits it.
  • Conditional loss weight wcond = 100
    Algorithm 2; tuned on validation ('turned out to be a reliable choice'); without it the model ignores dobs according to Section 6.1.
  • Unconditional training probability puncond = 0.5
    Algorithm 2; hand-set, needed for the classifier-free-guidance style training.
  • Guidance scale eta (sampling) = 0 to 1, user-controlled
    Algorithm 3; Section 6.3 shows inverse correlation with output diversity.
assumptions (5)
  • standard math Analytic posterior q(xt|x0,x1) in Eq. (17) is Gaussian with closed-form mean and variance under f=0 and Dirac-delta data (from Liu et al. [25]).
    Used by Algorithms 1-2 to sample noisy bridge states ct during training; borrowed from cited I2SB paper, Section 2.5.
  • standard math The amortized conditional SB problem (13) is equivalent to an extended-space SB (14), giving tractable pjoin and pjprior (from Shi et al. [24]).
    Grounds the conditional problem statement in Section 3, Eqs. (21)-(24).
  • domain assumption Test-time initial velocity models csmooth are generated from the same parametric family used in training: c1 = S_k(gamma z + (1-gamma) c0), with k in [8,16] and gamma in [0,0.2].
    Section 5.1.3: the operator 'does not have any basis in real-world data'; Table 6 shows degradation for shifted operators, so this assumption is load-bearing.
  • ad hoc to paper Joint training of conditional and unconditional drifts in one network, with zero-masking (puncond=0.5) and convex guidance combination (eta), produces valid posterior sampling control.
    Algorithms 2-3; the paper defers rigorous theory ('we leave rigorous theoretical treatment of the issue for future studies', Section 3).
  • domain assumption Fixed acquisition geometry and constant seismogram conditioning shape (5x64x64) across training and test.
    All experiments resize seismograms to a fixed shape; Section 6.4 flags this as a limitation for transferability.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Acoustic Waveform Inversion with Image-to-Image Schr\"odinger Bridges." pith.science (2026). https://pith.science/paper/TR5VXVY7

@misc{pith2026250615346,
  author       = {Pith},
  title        = {Pith review of: Acoustic Waveform Inversion with Image-to-Image Schr\"odinger Bridges},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TR5VXVY7}},
  note         = {Machine review of arXiv:2506.15346}
}
abstract

Recent developments in application of deep learning models to acoustic Full Waveform Inversion (FWI) are marked by the use of diffusion models as prior distributions for Bayesian-like inference procedures. The advantage of these methods is the ability to generate high-resolution samples, which are otherwise unattainable with classical inversion methods or other deep learning-based solutions. However, the iterative and stochastic nature of sampling from diffusion models along with heuristic nature of output control remain limiting factors for their applicability. For instance, an optimal way to include the approximate velocity model into diffusion-based inversion scheme remains unclear, even though it is considered an essential part of FWI pipeline. We address the issue by employing a Schr\"odinger Bridge that interpolates between the distributions of ground truth and smoothed velocity models. To facilitate the learning of nonlinear drifts that transfer samples between distributions we extend the concept of Image-to-Image Schr\"odinger Bridge ($\text{I}^2\text{SB}$) to conditional sampling, resulting in a conditional Image-to-Image Schr\"odinger Bridge (c$\text{I}^2\text{SB}$) framework. To validate our method, we assess its effectiveness in reconstructing the reference velocity model from its smoothed approximation, coupled with the observed seismic signal of fixed shape. Our experiments demonstrate that the proposed solution outperforms our reimplementation of conditional diffusion model suggested in earlier works, while requiring only a few neural function evaluations (NFEs) to achieve sample fidelity superior to that attained with supervised learning-based approach. The supplementary code implementing the algorithms described in this paper can be found in the repository https://github.com/stankevich-mipt/seismic_inversion_via_I2SB.

Figures

Figures reproduced from arXiv: 2506.15346 by the authors.

Figure 1
Figure 1. Conditional Image-to-Image Schr¨odinger Bridge in application to acoustic seismic inversion [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. A schematic of attention-augmented UNet network used for numerical studies throughout the study. Original architecture was [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. OpenFWI velocity map and seismic data samples [26] [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Initial guess distributions obtained via application of degradation operator to velocity model samples from different datasets [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Velocity models obtained through deep-learning based seismic inversion in comparison with ground truth velocity models. The [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Velocity models obtained through deep-learning based seismic inversion in comparison with ground truth velocity models. The [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Generative process of cI2SB for model instances trained on the entirety of the OpenFWI collection. 6 Discussion 6.1 Objective reweighting The use of sampling scheme 3 for ci2sb-seismic-small model instance trained with alg. 2 while wcond is set 0 leads to the inferior …
Figure 8
Figure 8. Figure 8: Performance comparison for introduced frameworks with different sets of hyperparameters. Setting a higher value of loss [PITH_FULL_IMAGE:figures/full_fig_p018_8.png]
Figure 9
Figure 9. Figure 9: Velocity models reconstructed with cI2SB for varying number of NFEs 19 [PITH_FULL_IMAGE:figures/full_fig_p019_9.png]
Figure 10
Figure 10. Figure 10: Reconstructed velocity model slices for varying number of NFEs [PITH_FULL_IMAGE:figures/full_fig_p020_10.png]
Figure 11
Figure 11. Figure 11: Dependency of validation metrics from timestep [PITH_FULL_IMAGE:figures/full_fig_p020_11.png]
Figure 12
Figure 12. Figure 12: Velocity model reconstructed with cSGM for varying number of NFEs [PITH_FULL_IMAGE:figures/full_fig_p021_12.png]
Figure 13
Figure 13. Figure 13: Velocity models reconstructed with cI2SB for varying values of guidance scale 22 [PITH_FULL_IMAGE:figures/full_fig_p022_13.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 28 canonical work pages

  1. [1]

    B. H. Russell, Introduction to seismic inversion methods . SEG Books, 1988

    work page 1988

  2. [2]

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

    work page 2017

  3. [3]

    J. O. Robertsson, B. Bednar, J. Blanch, C. Kostov, and D.-J. van Manen, Introduction to the supplement on seismic modeling with applications to acquisition, processing, and interpretation , 2007

    work page 2007

  4. [4]

    Accuracy of finite-difference modeling of the acoustic wave equa- tion,

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

    work page 1974

  5. [5]

    Deep learning for seismic inverse problems: Toward the accel- eration of geophysical analysis workflows,

    A. Adler, M. Araya-Polo, and T. Poggio, “Deep learning for seismic inverse problems: Toward the accel- eration of geophysical analysis workflows,” IEEE Signal Processing Magazine , vol. 38, no. 2, pp. 89–119, 2021

    work page 2021

  6. [6]

    Applications of deep neural networks in exploration seismology: A technical survey,

    S. M. Mousavi, G. C. Beroza, T. Mukerji, and M. Rasht-Behesht, “Applications of deep neural networks in exploration seismology: A technical survey,” Geophysics, vol. 89, no. 1, WA95–WA115, 2024

    work page 2024

  7. [7]

    Deep-learning tomography,

    M. Araya-Polo, J. Jennings, A. Adler, and T. Dahlke, “Deep-learning tomography,” The Leading Edge , vol. 37, no. 1, pp. 58–66, 2018. 15

    work page 2018

  8. [8]

    Adjoint-driven deep-learning seismic full-waveform inversion,

    W. Zhang, J. Gao, Z. Gao, and H. Chen, “Adjoint-driven deep-learning seismic full-waveform inversion,” IEEE Transactions on Geoscience and Remote Sensing , vol. 59, no. 10, pp. 8913–8932, 2020

    work page 2020

Show all 48 references
  1. [9]

    Deep-learning full-waveform inversion using seismic migration images,

    W. Zhang and J. Gao, “Deep-learning full-waveform inversion using seismic migration images,” IEEE Transactions on Geoscience and Remote Sensing , vol. 60, pp. 1–18, 2021

    2021

  2. [10]

    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, 2019

    2019

  3. [11]

    Deep-learning inversion: A next-generation seismic velocity model building method,

    F. Yang and J. Ma, “Deep-learning inversion: A next-generation seismic velocity model building method,” Geophysics, vol. 84, no. 4, R583–R599, 2019

    2019

  4. [12]

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

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

    2020

  5. [13]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in neural information processing systems, vol. 33, pp. 6840–6851, 2020

    2020

  6. [14]

    Diffusion models: A comprehensive survey of methods and applica- tions,

    L. Yang, Z. Zhang, Y. Song, et al., “Diffusion models: A comprehensive survey of methods and applica- tions,” ACM Computing Surveys , vol. 56, no. 4, pp. 1–39, 2023

    2023

  7. [15]

    Score-based generative modeling through stochastic differential equations,

    Y. Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole, “Score-based generative modeling through stochastic differential equations,” arXiv preprint arXiv:2011.13456 , 2020

    2011 arXiv

  8. [16]

    Srdiff: Single image super-resolution with diffusion probabilistic models,

    H. Li, Y. Yang, M. Chang, et al., “Srdiff: Single image super-resolution with diffusion probabilistic models,” Neurocomputing, vol. 479, pp. 47–59, 2022

    2022

  9. [17]

    Repaint: Inpainting using denoising diffusion probabilistic models,

    A. Lugmayr, M. Danelljan, A. Romero, F. Yu, R. Timofte, and L. Van Gool, “Repaint: Inpainting using denoising diffusion probabilistic models,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 11 461–11 471

    2022

  10. [18]

    Denoising diffusion restoration models,

    B. Kawar, M. Elad, S. Ermon, and J. Song, “Denoising diffusion restoration models,” Advances in Neural Information Processing Systems, vol. 35, pp. 23 593–23 606, 2022

    2022

  11. [19]

    Ho and T

    J. Ho and T. Salimans, Classifier-free diffusion guidance, 2022. arXiv: 2207.12598 [cs.LG]

    2022 arXiv

  12. [20]

    Rombach, A

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, High-resolution image synthesis with latent diffusion models , 2022. arXiv: 2112.10752 [cs.CV]

    2022 arXiv

  13. [21]

    F. Wang, X. Huang, and T. A. Alkhalifah, IEEE Transactions on Geoscience and Remote Sensing, vol. 61, pp. 1–11, 2023

    2023

  14. [22]

    Controllable seismic velocity synthesis using generative dif- fusion models,

    F. Wang, X. Huang, and T. Alkhalifah, “Controllable seismic velocity synthesis using generative dif- fusion models,” Journal of Geophysical Research: Machine Learning and Computation , vol. 1, no. 3, e2024JH000153, 2024

    2024

  15. [23]

    Diffusionvel: Multi-information integrated velocity inversion using gen- erative diffusion models,

    H. Zhang, Y. Li, and J. Huang, “Diffusionvel: Multi-information integrated velocity inversion using gen- erative diffusion models,” arXiv preprint arXiv:2410.21776 , 2024

    2024 arXiv

  16. [24]

    Conditional simulation using diffusion schr¨ odinger bridges,

    Y. Shi, V. De Bortoli, G. Deligiannidis, and A. Doucet, “Conditional simulation using diffusion schr¨ odinger bridges,” in Uncertainty in Artificial Intelligence , PMLR, 2022, pp. 1792–1802

    2022

  17. [25]

    G.-H. Liu, A. Vahdat, D.-A. Huang, E. A. Theodorou, W. Nie, and A. Anandkumar, I2sb: Image-to-image schr¨ odinger bridge, 2023. arXiv: 2302.05872 [cs.CV]

    2023 arXiv

  18. [26]

    Openfwi: Benchmark seismic datasets for machine learning-based full waveform inversion,

    C. Deng, Y. Feng, S. Feng, et al., “Openfwi: Benchmark seismic datasets for machine learning-based full waveform inversion,” CoRR, vol. abs/2111.02926, 2021. arXiv: 2111.02926

    2021 arXiv

  19. [27]

    An overview of full-waveform inversion in exploration geophysics,

    “An overview of full-waveform inversion in exploration geophysics,” Geophysics, vol. 74, no. 6, WCC1– WCC26, 2009

    2009

  20. [28]

    Reverse-time diffusion equation models,

    B. D. Anderson, “Reverse-time diffusion equation models,” Stochastic Processes and their Applications , vol. 12, no. 3, pp. 313–326, 1982

    1982

  21. [29]

    A connection between score matching and denoising autoencoders,

    P. Vincent, “A connection between score matching and denoising autoencoders,” Neural computation, vol. 23, no. 7, pp. 1661–1674, 2011

    2011

  22. [30]

    Sur la th´ eorie relativiste de l’´ electron et l’interpr´ etation de la m´ ecanique quantique,

    E. Schr¨ odinger, “Sur la th´ eorie relativiste de l’´ electron et l’interpr´ etation de la m´ ecanique quantique,” in Annales de l’institut Henri Poincar´ e, vol. 2, 1932, pp. 269–310

    1932

  23. [31]

    On free energy, stochastic control, and schr¨ odinger processes,

    M. Pavon and A. Wakolbinger, “On free energy, stochastic control, and schr¨ odinger processes,” in Mod- eling, Estimation and Control of Systems with Uncertainty: Proceedings of a Conference held in Sopron, Hungary, September 1990 , Springer, 1991, pp. 334–348

    1990

  24. [32]

    A survey of the schr\

    C. L´ eonard, “A survey of the schr\” odinger problem and some of its connections with optimal transport,” arXiv preprint arXiv:1308.0215 , 2013. 16

    2013 arXiv

  25. [33]

    Stochastic control liaisons: Richard sinkhorn meets gaspard monge on a schrodinger bridge,

    Y. Chen, T. T. Georgiou, and M. Pavon, “Stochastic control liaisons: Richard sinkhorn meets gaspard monge on a schrodinger bridge,” Siam Review, vol. 63, no. 2, pp. 249–313, 2021

    2021

  26. [34]

    Likelihood training of schr \

    T. Chen, G.-H. Liu, and E. A. Theodorou, “Likelihood training of schr \” odinger bridge using forward- backward sdes theory,” arXiv preprint arXiv:2110.11291 , 2021

    2021 arXiv

  27. [35]

    Diffusion schr¨ odinger bridge with applications to score-based generative modeling,

    V. De Bortoli, J. Thornton, J. Heng, and A. Doucet, “Diffusion schr¨ odinger bridge with applications to score-based generative modeling,” Advances in Neural Information Processing Systems, vol. 34, pp. 17 695– 17 709, 2021

    2021

  28. [36]

    Y. Chen, T. T. Georgiou, and M. Pavon, Stochastic control liaisons: Richard sinkhorn meets gaspard monge on a schroedinger bridge , 2020. arXiv: 2005.10963 [math.OC]

    2020 arXiv

  29. [37]

    Large scale gan training for high fidelity natural image synthe- sis,

    A. Brock, J. Donahue, and K. Simonyan, “Large scale gan training for high fidelity natural image synthe- sis,” arXiv preprint arXiv:1809.11096 , 2018

    2018 arXiv

  30. [38]

    Chung, B

    H. Chung, B. Sim, D. Ryu, and J. C. Ye, Improving diffusion models for inverse problems using manifold constraints, 2022. arXiv: 2206.00941 [cs.LG]

    2022 arXiv

  31. [39]

    Diffusion models beat gans on image synthesis,

    P. Dhariwal and A. Nichol, “Diffusion models beat gans on image synthesis,” CoRR, vol. abs/2105.05233,

    arXiv

  32. [40]

    Karras, M

    T. Karras, M. Aittala, T. Aila, and S. Laine, Elucidating the design space of diffusion-based generative models, 2022. arXiv: 2206.00364 [cs.CV]

    2022 arXiv

  33. [41]

    Improved denoising diffusion probabilistic models,

    A. Nichol and P. Dhariwal, “Improved denoising diffusion probabilistic models,” CoRR, vol. abs/2102.09672,

    arXiv

  34. [42]

    Denoising diffusion implicit models,

    J. Song, C. Meng, and S. Ermon, “Denoising diffusion implicit models,” arXiv preprint arXiv:2010.02502, 2020

    2010 arXiv

  35. [43]

    Exponential convergence of langevin distributions and their discrete approximations,

    G. O. Roberts and R. L. Tweedie, “Exponential convergence of langevin distributions and their discrete approximations,” Bernoulli, pp. 341–363, 1996

    1996

  36. [44]

    Plug & play generative networks: Conditional iterative generation of images in latent space,

    A. Nguyen, J. Clune, Y. Bengio, A. Dosovitskiy, and J. Yosinski, “Plug & play generative networks: Conditional iterative generation of images in latent space,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 4467–4477

    2017

  37. [45]

    Diffusion posterior sampling for general noisy inverse problems,

    H. Chung, J. Kim, M. T. Mccann, M. L. Klasky, and J. C. Ye, “Diffusion posterior sampling for general noisy inverse problems,” arXiv preprint arXiv:2209.14687 , 2022

    2022 arXiv

  38. [46]

    An empirical study of large-scale data-driven full waveform inversion,

    P. Jin, Y. Feng, S. Feng, et al., “An empirical study of large-scale data-driven full waveform inversion,” Scientific Reports, vol. 14, no. 1, p. 20 034, 2024

    2024

  39. [47]

    The perception-distortion tradeoff,

    Y. Blau and T. Michaeli, “The perception-distortion tradeoff,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 6228–6237

    2018

  40. [48]

    Inversion by direct iteration: An alternative to denoising diffusion for image restoration,

    M. Delbracio and P. Milanfar, “Inversion by direct iteration: An alternative to denoising diffusion for image restoration,” arXiv preprint arXiv:2303.11435 , 2023. 17 Supplementary figures (a) (b) (c) Figure 8: Performance comparison for introduced frameworks with different se...

    2023 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.