pith. machine review for the scientific record. sign in

arxiv: 2605.14370 · v1 · submitted 2026-05-14 · ⚛️ physics.geo-ph · cs.AI· physics.comp-ph

Recognition: 2 theorem links

· Lean Theorem

Deciphering Neural Reparameterized Full-Waveform Inversion with Neural Sensitivity Kernel and Wave Tangent Kernel

Ruihua Chen , Yisi Luo , Bangyu Wu , Xile Zhao , Deyu Meng
Authors on Pith no claims yet

Pith reviewed 2026-05-15 01:54 UTC · model grok-4.3

classification ⚛️ physics.geo-ph cs.AIphysics.comp-ph
keywords neural reparameterized full-waveform inversionneural sensitivity kernelwave tangent kernelneural tangent kernelspectral filteringconvergence analysisseismic inversionmedical imaging
0
0 comments X

The pith

The neural tangent kernel from neural reparameterization modulates sensitivity and wave tangent kernels in full-waveform inversion, producing spectral filtering and wavenumber shifts that govern convergence.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper introduces the neural sensitivity kernel and wave tangent kernel to analyze why neural reparameterization makes full-waveform inversion more robust to bad initial models while slowing high-resolution recovery. The central mechanism is the neural tangent kernel induced by the network, which adaptively changes the original kernels and creates spectral filtering, gradient wavenumber modulation, and wave frequency bias. These changes tie the inversion's convergence speed and resolution directly to the eigen-structures of the modulated kernels. The authors then design new neural reparameterizations that target better eigen-structures and demonstrate gains on seismic data and the first extension to medical imaging.

Core claim

The neural tangent kernel induced by neural representation adaptively modulates the original sensitivity kernel and wave tangent kernel. This modulation produces a spectral filtering effect, gradient wavenumber modulation, and wave frequency bias, directly connecting the convergence behavior of neural reparameterized full-waveform inversion to the eigen-structures of the neural sensitivity kernel and wave tangent kernel.

What carries the argument

Neural sensitivity kernel (NSK) and wave tangent kernel (WTK), which the neural tangent kernel (NTK) from the neural network modulates to control convergence in both model and data domains.

If this is right

  • Enhanced neural reparameterizations with tailored eigen-structures in NSK and WTK improve both inversion accuracy and computational efficiency.
  • The spectral filtering effect explains reduced dependence on high-quality initial models in NeurFWI.
  • Wave frequency bias from the modulation limits high-resolution recovery unless the eigen-structures are adjusted.
  • The same NSK and WTK framework applies beyond seismic exploration to medical imaging applications.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Network architecture choices that shape the NTK spectrum could be used to tune inversion resolution without hand-crafted regularization.
  • The modulation mechanism may generalize to other neural-reparameterized inverse problems such as electromagnetic tomography.
  • Explicit control of kernel eigen-structures offers a route to faster high-wavenumber recovery in data-limited settings.

Load-bearing premise

The modulation of the original kernels by the neural tangent kernel connects directly to convergence rates through eigen-structure analysis without hidden approximations in the derivation.

What would settle it

A controlled experiment in which the neural network architecture is changed to alter the neural tangent kernel spectrum while keeping other factors fixed, yet the predicted shifts in gradient spectra, frequency content, or convergence speed fail to appear.

read the original abstract

Full-waveform inversion (FWI) estimates unknown parameters in the wave equation from limited boundary measurements. Recent advances in neural reparameterized FWI (NeurFWI) demonstrate that representing the parameters using a neural network can reduce the reliance on the high-quality initial model and wavefield data, at the cost of slow high-resolution convergence. However, its underlying theoretical mechanism remains unclear. In this study, we establish the neural sensitivity kernel (NSK) and the wave tangent kernel (WTK) to analyze their convergence behavior from both model and data domains. These theoretical frameworks show that the neural tangent kernel (NTK) induced by neural representation adaptively modulates the original sensitivity and wave tangent kernels. This modulation leads to several key outcomes, i.e., the spectral filtering effect, the gradient wavenumber modulation, and the wave frequency bias, connecting the convergence behavior of NeurFWI with the eigen-structures of NSK and WTK. Building on these insights, we propose several enhanced NeurFWI methods with tailored eigen-structures in NSK and WTK to improve inversion performances and efficiency. We numerically validate these theoretical claims and the proposed methods in seismic exploration, and firstly extend their application to medical imaging.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper claims to decipher the convergence behavior of neural reparameterized full-waveform inversion (NeurFWI) by establishing neural sensitivity kernel (NSK) and wave tangent kernel (WTK). It shows that the neural tangent kernel (NTK) adaptively modulates the original sensitivity and wave tangent kernels, leading to spectral filtering effect, gradient wavenumber modulation, and wave frequency bias. These are connected to the eigen-structures of NSK and WTK, and enhanced NeurFWI methods are proposed and numerically validated in seismic and medical imaging applications.

Significance. If the theoretical links hold, this provides a novel framework for understanding and improving NeurFWI by tailoring kernel eigen-structures, potentially reducing reliance on initial models while addressing slow convergence. The extension to medical imaging broadens the impact, and the numerical validation offers practical evidence, though the strength depends on rigorous proof of the modulation effects without hidden assumptions.

major comments (2)
  1. The adaptive modulation by NTK is key to the spectral filtering and other effects, but the manuscript must explicitly derive how the NTK alters the eigen-structures of NSK and WTK (e.g., in the section introducing these kernels) to confirm no circularity or unstated approximations like Born linearization are involved.
  2. The connection between the modulated kernels' eigen-structures and the observed convergence behavior (slow high-resolution) needs to be supported by specific analysis or theorems; without this, the explanation for NeurFWI's properties remains incomplete.
minor comments (2)
  1. The abstract is information-dense; expanding slightly on the proposed enhanced methods would improve clarity for readers.
  2. Include more details on the specific metrics used to demonstrate improvements in the enhanced methods, such as convergence curves or resolution measures.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thorough review and valuable comments. We address each major comment below and agree to make revisions to strengthen the theoretical derivations and connections as suggested.

read point-by-point responses

  1. Referee: The adaptive modulation by NTK is key to the spectral filtering and other effects, but the manuscript must explicitly derive how the NTK alters the eigen-structures of NSK and WTK (e.g., in the section introducing these kernels) to confirm no circularity or unstated approximations like Born linearization are involved.

    Authors: In the manuscript, the NSK and WTK are introduced in Section 3.2 and 3.3, where we derive the modulated kernels explicitly using the chain rule: the gradient with respect to model parameters involves the NTK matrix multiplying the original sensitivity kernel. This leads to the eigen-structures being altered by the NTK's spectral properties without circularity, as the NTK is computed from the network architecture independently. No Born approximation is used; the derivation holds for the nonlinear forward operator via the tangent linearization at each iteration. To address the concern, we will expand the derivation in a new subsection 3.2.1 with explicit steps showing the operator modulation and its effect on eigenvalues, including a proof that the modulation is adaptive and non-circular. revision: yes

  2. Referee: The connection between the modulated kernels' eigen-structures and the observed convergence behavior (slow high-resolution) needs to be supported by specific analysis or theorems; without this, the explanation for NeurFWI's properties remains incomplete.

    Authors: We provide the connection through the eigenvalue analysis in Section 4, where we show that the NTK modulation damps high-wavenumber eigenvalues, leading to slower convergence for high-resolution features. This is supported by the spectral decomposition and numerical experiments. To make it more rigorous, we will add a theorem in the revised manuscript that formalizes the convergence rate as a function of the smallest eigenvalues of the modulated WTK/NSK, with a proof sketch based on the gradient descent dynamics in the kernel space. revision: yes

Circularity Check

1 steps flagged

NTK modulation of NSK/WTK is definitional via neural reparameterization; eigen-structure claims reduce to built-in properties

specific steps
  1. self definitional [Abstract]
    "These theoretical frameworks show that the neural tangent kernel (NTK) induced by neural representation adaptively modulates the original sensitivity and wave tangent kernels. This modulation leads to several key outcomes, i.e., the spectral filtering effect, the gradient wavenumber modulation, and the wave frequency bias, connecting the convergence behavior of NeurFWI with the eigen-structures of NSK and WTK."

    NSK and WTK are introduced as the sensitivity and tangent kernels under neural reparameterization; the 'adaptive modulation' by NTK is therefore part of the definition rather than a derived property. The claimed connection between eigen-structures and convergence behavior is then asserted from these same constructed kernels, making the central theoretical outcome equivalent to the input definitions.

full rationale

The paper defines NSK and WTK explicitly in terms of the neural representation and its induced NTK, then asserts that this modulation produces spectral filtering, wavenumber modulation, and frequency bias whose eigen-structures control convergence. This link is presented as an analysis result but follows directly from the construction of the kernels rather than an independent derivation. No self-citations or fitted predictions are load-bearing; the circularity is limited to the self-definitional framing of the modulation effect.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

The central claims rest on the assumption that neural reparameterization induces an NTK that modulates physical kernels in the described manner; no explicit free parameters or invented entities beyond the new kernels are stated in the abstract.

axioms (1)
  • domain assumption Neural network representation of model parameters induces a neural tangent kernel that adaptively modulates sensitivity and wave tangent kernels.
    This is the core premise used to derive the spectral filtering, wavenumber modulation, and frequency bias effects.
invented entities (2)
  • Neural sensitivity kernel (NSK) no independent evidence
    purpose: Analyze convergence behavior in the model domain for NeurFWI.
    Newly introduced analytical tool.
  • Wave tangent kernel (WTK) no independent evidence
    purpose: Analyze convergence behavior in the data domain for NeurFWI.
    Newly introduced analytical tool.

pith-pipeline@v0.9.0 · 5535 in / 1272 out tokens · 35571 ms · 2026-05-15T01:54:14.823698+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

174 extracted references · 174 canonical work pages · 1 internal anchor

  1. [1]

    Geophysics , volume=

    An overview of full-waveform inversion in exploration geophysics , author=. Geophysics , volume=. 2009 , publisher=

    work page 2009

  2. [2]

    International Conference on Computational Learning Theory , pages=

    Mercer’s theorem, feature maps, and smoothing , author=. International Conference on Computational Learning Theory , pages=. 2006 , organization=

    work page 2006

  3. [3]

    Geophysical Prospecting , volume=

    Full waveform inversion based on inversion network reparameterized velocity , author=. Geophysical Prospecting , volume=. 2023 , publisher=

    work page 2023

  4. [4]

    Geophysics , volume=

    Reparameterized full-waveform inversion using deep neural networks , author=. Geophysics , volume=. 2021 , publisher=

    work page 2021

  5. [5]

    Topology and its Applications , volume=

    A Fubini theorem , author=. Topology and its Applications , volume=. 2007 , publisher=

    work page 2007

  6. [6]

    Journal of Geophysical Research: Machine Learning and Computation , volume=

    Learned regularizations for multi-parameter elastic full waveform inversion using diffusion models , author=. Journal of Geophysical Research: Machine Learning and Computation , volume=. 2024 , publisher=

    work page 2024

  7. [7]

    Geophysical Journal International , volume=

    Compensating for source and receiver ghost effects in full waveform inversion and reverse time migration for marine streamer data , author=. Geophysical Journal International , volume=. 2015 , publisher=

    work page 2015

  8. [8]

    Geophysics , volume=

    Anisotropic 3D full-waveform inversion , author=. Geophysics , volume=. 2013 , publisher=

    work page 2013

  9. [9]

    Geophysics , volume=

    Waveform inversion of seismic first arrivals acquired on irregular surface , author=. Geophysics , volume=. 2022 , publisher=

    work page 2022

  10. [10]

    Geophysics , volume=

    Effects of surface scattering in full-waveform inversion , author=. Geophysics , volume=. 2009 , publisher=

    work page 2009

  11. [11]

    Geophysical Journal International , volume=

    Theoretical background for continental-and global-scale full-waveform inversion in the time--frequency domain , author=. Geophysical Journal International , volume=. 2008 , publisher=

    work page 2008

  12. [12]

    Geophysical Journal International , volume=

    Two-dimensional velocity models from wide-angle seismic data by wavefield inversion , author=. Geophysical Journal International , volume=. 1996 , publisher=

    work page 1996

  13. [13]

    IEEE Geoscience and Remote Sensing Letters , author =

    Optimal. IEEE Geoscience and Remote Sensing Letters , author =. 2025 , pages =. doi:10.1109/LGRS.2025.3550682 , abstract =

    work page doi:10.1109/lgrs.2025.3550682 2025

  14. [14]

    and Gholami, A

    Aghazade, K. and Gholami, A. and S.Aghamiry, H. , year =. Full. 84th. doi:10.3997/2214-4609.202310431 , abstract =

    work page doi:10.3997/2214-4609.202310431

  15. [15]

    2006 , publisher=

    Inverse problems for partial differential equations , author=. 2006 , publisher=

    work page 2006

  16. [16]

    Geophysics , volume=

    Multiscale seismic waveform inversion , author=. Geophysics , volume=. 1995 , publisher=

    work page 1995

  17. [17]

    Geophysics , volume=

    Inversion of seismic reflection data in the acoustic approximation , author=. Geophysics , volume=. 1984 , publisher=

    work page 1984

  18. [18]

    SIAM Journal on Scientific Computing , volume=

    Full waveform inversion guided by travel time tomography , author=. SIAM Journal on Scientific Computing , volume=. 2017 , publisher=

    work page 2017

  19. [19]

    SIAM Journal on Scientific Computing , volume=

    A time-domain preconditioned truncated Newton approach to visco-acoustic multiparameter full waveform inversion , author=. SIAM Journal on Scientific Computing , volume=. 2018 , publisher=

    work page 2018

  20. [20]

    Geophysical Journal International , volume=

    The optimized gradient method for full waveform inversion and its spectral implementation , author=. Geophysical Journal International , volume=. 2016 , publisher=

    work page 2016

  21. [21]

    Journal of Geophysical Research: Solid Earth , volume=

    Separation of migration and tomography modes of full-waveform inversion in the plane wave domain , author=. Journal of Geophysical Research: Solid Earth , volume=. 2018 , publisher=

    work page 2018

  22. [22]

    Geophysical Journal International , pages=

    Implicit Full Waveform Inversion with Adaptive Fourier Frequency Bases Learning , author=. Geophysical Journal International , pages=. 2025 , publisher=

    work page 2025

  23. [23]

    Journal of Geophysics and Engineering , volume=

    Waveform inversion with structural regularizing constraint based on gradient decomposition , author=. Journal of Geophysics and Engineering , volume=. 2024 , publisher=

    work page 2024

  24. [24]

    SIAM review , volume=

    Tensor decompositions and applications , author=. SIAM review , volume=. 2009 , publisher=

    work page 2009

  25. [25]

    2012 , publisher=

    Matrix analysis , author=. 2012 , publisher=

    work page 2012

  26. [26]

    SEG 2014 benchmark data , author=

    work page 2014

  27. [27]

    Geophysics , volume=

    Efficient waveform inversion and imaging: A strategy for selecting temporal frequencies , author=. Geophysics , volume=. 2004 , publisher=

    work page 2004

  28. [28]

    Advances in Neural Information Processing Systems , volume=

    The shaped transformer: Attention models in the infinite depth-and-width limit , author=. Advances in Neural Information Processing Systems , volume=

    work page

  29. [29]

    Journal of Geophysical Research: Machine Learning and Computation , volume=

    How Does Neural Network Reparametrization Improve Geophysical Inversion? , author=. Journal of Geophysical Research: Machine Learning and Computation , volume=. 2025 , publisher=

    work page 2025

  30. [30]

    IEEE Transactions on Industrial Electronics , number=

    An extension of Parseval's theorem and its use in calculating transient energy in the frequency domain , author=. IEEE Transactions on Industrial Electronics , number=. 2007 , publisher=

    work page 2007

  31. [31]

    Geophysical Journal International , volume=

    Implicit multiparameter full waveform inversion of multioffset ground penetrating radar data , author=. Geophysical Journal International , volume=. 2025 , publisher=

    work page 2025

  32. [32]

    Science China Mathematics , volume=

    A comparative analysis of optimization and generalization properties of two-layer neural network and random feature models under gradient descent dynamics , author=. Science China Mathematics , volume=. 2020 , publisher=

    work page 2020

  33. [33]

    IEEE Transactions on Geoscience and Remote Sensing , year=

    Enhancing Tomography Component of Full-Waveform Inversion Based on Gradient Decomposition , author=. IEEE Transactions on Geoscience and Remote Sensing , year=

    work page

  34. [34]

    Inverse Problems , volume=

    Inverse problems for abstract evolution equations with applications in electrodynamics and elasticity , author=. Inverse Problems , volume=. 2016 , publisher=

    work page 2016

  35. [35]

    Inverse Problems , volume=

    Tangential cone condition and Lipschitz stability for the full waveform forward operator in the acoustic regime , author=. Inverse Problems , volume=. 2021 , publisher=

    work page 2021

  36. [36]

    SEG international exposition and annual meeting , pages=

    Deep learning prior models from seismic images for full-waveform inversion , author=. SEG international exposition and annual meeting , pages=. 2017 , organization=

    work page 2017

  37. [37]

    IEEE Transactions on Geoscience and Remote Sensing , volume=

    Deep-learning full-waveform inversion using seismic migration images , author=. IEEE Transactions on Geoscience and Remote Sensing , volume=. 2021 , publisher=

    work page 2021

  38. [38]

    SIAM Journal on Applied Mathematics , volume=

    Level set--based shape optimization approach for sharp-interface reconstructions in time-domain full waveform inversion , author=. SIAM Journal on Applied Mathematics , volume=. 2021 , publisher=

    work page 2021

  39. [39]

    SIAM Journal on Scientific Computing , volume=

    Full waveform inversion using extended and simultaneous sources , author=. SIAM Journal on Scientific Computing , volume=. 2021 , publisher=

    work page 2021

  40. [40]

    SIAM Journal on Imaging Sciences , volume=

    Waveform inversion with a data driven estimate of the internal wave , author=. SIAM Journal on Imaging Sciences , volume=. 2023 , publisher=

    work page 2023

  41. [41]

    SIAM Journal on Scientific Computing , volume=

    Seismic tomography with random batch gradient reconstruction , author=. SIAM Journal on Scientific Computing , volume=. 2023 , publisher=

    work page 2023

  42. [42]

    SIAM Journal on Imaging Sciences , volume=

    Complex-valued imaging with total variation regularization: an application to full-waveform inversion in visco-acoustic media , author=. SIAM Journal on Imaging Sciences , volume=. 2021 , publisher=

    work page 2021

  43. [43]

    SIAM Journal on Applied Mathematics , volume=

    Inverse problems for abstract evolution equations II: higher order differentiability for viscoelasticity , author=. SIAM Journal on Applied Mathematics , volume=. 2019 , publisher=

    work page 2019

  44. [44]

    Natural Hazards , volume=

    Fire danger monitoring using ERS-1 SAR images in the case of northern boreal forests , author=. Natural Hazards , volume=. 2002 , publisher=

    work page 2002

  45. [45]

    2021 , publisher=

    Introduction to inverse problems for differential equations , author=. 2021 , publisher=

    work page 2021

  46. [46]

    SEG Technical Program Expanded Abstracts 2017 , pages=

    A stochastic L-BFGS approach for full-waveform inversion , author=. SEG Technical Program Expanded Abstracts 2017 , pages=. 2017 , publisher=

    work page 2017

  47. [47]

    SIAM review , volume=

    Full waveform inversion and the truncated Newton method , author=. SIAM review , volume=. 2017 , publisher=

    work page 2017

  48. [48]

    Pure and Applied Geophysics , volume=

    Effects of conjugate gradient methods and step-length formulas on the multiscale full waveform inversion in time domain: Numerical experiments , author=. Pure and Applied Geophysics , volume=. 2017 , publisher=

    work page 2017

  49. [49]

    GEOPHYSICS , volume =

    Fuqiang Chen and Daniel Peter and Matteo Ravasi , title =. GEOPHYSICS , volume =. 2022 , doi =

    work page 2022

  50. [50]

    Geophysics , volume=

    Review of Physics-Informed Machine Learning Inversion of Geophysical Data , author=. Geophysics , volume=. 2024 , publisher=

    work page 2024

  51. [51]

    IEEE Transactions on Computational Imaging , volume=

    InversionNet: An efficient and accurate data-driven full waveform inversion , author=. IEEE Transactions on Computational Imaging , volume=. 2019 , publisher=

    work page 2019

  52. [52]

    Journal of Geophysical Research: Solid Earth , volume=

    Physics-informed neural networks (PINNs) for wave propagation and full waveform inversions , author=. Journal of Geophysical Research: Solid Earth , volume=. 2022 , publisher=

    work page 2022

  53. [53]

    Geophysics , volume=

    Mapping full seismic waveforms to vertical velocity profiles by deep learning , author=. Geophysics , volume=. 2021 , publisher=

    work page 2021

  54. [54]

    Journal of Computational Physics , volume=

    Self-adaptive physics-informed neural networks , author=. Journal of Computational Physics , volume=. 2023 , publisher=

    work page 2023

  55. [55]

    Advances in Neural Information Processing Systems , volume=

    The challenges of the nonlinear regime for physics-informed neural networks , author=. Advances in Neural Information Processing Systems , volume=

    work page

  56. [56]

    arXiv preprint arXiv:2412.05545 , year=

    Convergence analysis of wide shallow neural operators within the framework of Neural Tangent Kernel , author=. arXiv preprint arXiv:2412.05545 , year=

    work page arXiv

  57. [57]

    IEEE Transactions on Geoscience and Remote Sensing , year=

    Quanv4eo: empowering earth observation by means of quanvolutional neural networks , author=. IEEE Transactions on Geoscience and Remote Sensing , year=

    work page

  58. [58]

    arXiv preprint arXiv:2503.11029 , year=

    Neural Tangent Kernel of Neural Networks with Loss Informed by Differential Operators , author=. arXiv preprint arXiv:2503.11029 , year=

    work page arXiv

  59. [59]

    arXiv preprint arXiv:2412.17518 , year=

    Optimal convergence rates for neural operators , author=. arXiv preprint arXiv:2412.17518 , year=

    work page arXiv

  60. [60]

    arXiv preprint arXiv:2504.11130 , year=

    Divergence of empirical neural tangent kernel in classification problems , author=. arXiv preprint arXiv:2504.11130 , year=

    work page arXiv

  61. [61]

    Toward a better understanding of fourier neural operators from a spectral perspective,

    Toward a Better Understanding of Fourier Neural Operators from a Spectral Perspective , author=. arXiv preprint arXiv:2404.07200 , year=

    work page arXiv

  62. [62]

    Neural Processing Letters , volume=

    GPINN with neural tangent kernel technique for nonlinear two point boundary value problems , author=. Neural Processing Letters , volume=. 2024 , publisher=

    work page 2024

  63. [63]

    Unsupervised Learning of Full-Waveform Inversion: Connecting

    Peng Jin and Xitong Zhang and Yinpeng Chen and Sharon X Huang and Zicheng Liu and Youzuo Lin , booktitle=. Unsupervised Learning of Full-Waveform Inversion: Connecting

    work page

  64. [64]

    Journal of Computational physics , volume=

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations , author=. Journal of Computational physics , volume=. 2019 , publisher=

    work page 2019

  65. [65]

    IEEE Transactions on Geoscience and Remote Sensing , volume=

    Wavefield reconstruction inversion via physics-informed neural networks , author=. IEEE Transactions on Geoscience and Remote Sensing , volume=. 2021 , publisher=

    work page 2021

  66. [66]

    2019 IEEE Winter Conference on Applications of Computer Vision (WACV) , pages=

    VelocityGAN: Subsurface velocity image estimation using conditional adversarial networks , author=. 2019 IEEE Winter Conference on Applications of Computer Vision (WACV) , pages=. 2019 , organization=

    work page 2019

  67. [67]

    Advances in Neural Information Processing Systems , volume=

    On the linearity of large non-linear models: when and why the tangent kernel is constant , author=. Advances in Neural Information Processing Systems , volume=

    work page

  68. [68]

    Nature Reviews Earth & Environment , volume=

    Seismic wavefield imaging of Earth’s interior across scales , author=. Nature Reviews Earth & Environment , volume=. 2020 , publisher=

    work page 2020

  69. [69]

    IEEE Transactions on Geoscience and Remote Sensing , year=

    Full Waveform Inversion with Velocity Model Low-Rank Implicit Neural Representation , author=. IEEE Transactions on Geoscience and Remote Sensing , year=

    work page

  70. [70]

    Advances in neural information processing systems , volume=

    Fourier features let networks learn high frequency functions in low dimensional domains , author=. Advances in neural information processing systems , volume=

    work page

  71. [71]

    SIAM/ASA Journal on Uncertainty Quantification , volume=

    Stability of Gibbs posteriors from the Wasserstein loss for Bayesian full waveform inversion , author=. SIAM/ASA Journal on Uncertainty Quantification , volume=. 2021 , publisher=

    work page 2021

  72. [72]

    Geophysics , volume=

    ML-descent: An optimization algorithm for full-waveform inversion using machine learning , author=. Geophysics , volume=. 2020 , publisher=

    work page 2020

  73. [73]

    Archives of Computational Methods in Engineering , volume=

    Eighty years of the finite element method: Birth, evolution, and future , author=. Archives of Computational Methods in Engineering , volume=. 2022 , publisher=

    work page 2022

  74. [74]

    Waves and Imaging Class Notes--18.367 , author=

    work page

  75. [75]

    Geophysics , volume=

    Correlation-based reflection full-waveform inversion , author=. Geophysics , volume=. 2015 , publisher=

    work page 2015

  76. [76]

    Journal of Geophysics and Engineering , volume=

    Time domain full waveform inversion with low frequency wavefield decompression , author=. Journal of Geophysics and Engineering , volume=. 2018 , publisher=

    work page 2018

  77. [77]

    Geophysics , volume=

    Inversion= migration+ tomography , author=. Geophysics , volume=. 1989 , publisher=

    work page 1989

  78. [78]

    SIAM Journal on Mathematics of Data Science , volume=

    On the exact computation of linear frequency principle dynamics and its generalization , author=. SIAM Journal on Mathematics of Data Science , volume=. 2022 , publisher=

    work page 2022

  79. [79]

    Journal of Geophysics and Engineering , volume=

    Frequency-domain full waveform inversion with a scattering-integral approach and its sensitivity analysis , author=. Journal of Geophysics and Engineering , volume=. 2013 , publisher=

    work page 2013

  80. [80]

    2026 , issn =

    Fast ground penetrating radar dual-parameter full waveform inversion method accelerated by hybrid compilation of CUDA kernel function and PyTorch , journal =. 2026 , issn =. doi:https://doi.org/10.1016/j.cageo.2025.106101 , url =

    work page doi:10.1016/j.cageo.2025.106101 2026

Showing first 80 references.