Structure-dependent failure modes of neural priors in acoustic full-waveform inversion
Abstract
Frequency continuation does not improve every neural representation, and better velocity recovery need not imply better prediction of unseen waveforms. We compare a velocity grid with total-variation regularization, an adapted sinusoidal coordinate network (SIREN), and a frozen image generator with a full-grid residual in two-dimensional acoustic full-waveform inversion. All methods use the same frozen SWEEP discretization, acquisition, observations, background, and budget of 1,500 shot--model gradient evaluations. Three exposed synthetic targets represent inclined layering, curved layering, and a high-contrast curved interface drawn from a fault-labelled family. With 2-to-10 Hz continuation, the generative-residual model attains velocity RMSEs of 218.7, 108.0, and 341.9 m/s, compared with 232.5, 100.3, and 412.9 m/s for Grid+TV. Thus the latter wins on curved layering, while the former improves model RMSE by 5.9% and 17.2% on the other targets. Grid+TV nevertheless predicts held-out 10 Hz shots better on all three. SIREN is worse than the common background in every cell; continuation worsens two targets but improves one. A five-weight TV sensitivity study preserves the CurveFault model-error ranking. An equal-budget variational extension fails both recovery and interval coverage. These results establish conditional rankings, not universal superiority or an isolated spectral-bias mechanism. Full-grid residuals also prevent attributing gains uniquely to the generator. We advocate joint reporting of model recovery, unseen-shot prediction, optimization cost, and negative outcomes when auditing neural FWI.
Related
- PSDWII: Physical-Structure-Driven Waveform Inversion Imaging
- Structure-aware Bayesian full waveform inversion via diffusion-regularized Stein variational gradient descent
- Seismic imaging of the crust and upper mantle beneath Taiwan using full-waveform inversion of teleseismic body waves
- A comprehensive review of strategies to mitigate non-convexity in full waveform inversion
- Uncertainty Quantification in Full Waveform Inversion with Gradient Flows in Probability Space
BibTeX
@misc{yu2026structuredependent,
title = {{Structure-dependent failure modes of neural priors in acoustic full-waveform inversion}},
author = {Ziye Yu and Xin Liu and Yuqi Cai},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.31251},
eprint = {2609.31251},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.31251},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Structure-dependent failure modes of neural priors in acoustic full-waveform inversion
AU - Ziye Yu
AU - Xin Liu
AU - Yuqi Cai
PY - 2026
DA - 2026/09/25
PB - arXiv
DO - 10.48550/arxiv.2609.31251
UR - https://arxiv.org/abs/2609.31251
AB - Frequency continuation does not improve every neural representation, and better velocity recovery need not imply better prediction of unseen waveforms. We compare a velocity grid with total-variation regularization, an adapted sinusoidal coordinate network (SIREN), and a frozen image generator with a full-grid residual in two-dimensional acoustic full-waveform inversion. All methods use the same frozen SWEEP discretization, acquisition, observations, background, and budget of 1,500 shot--model gradient evaluations. Three exposed synthetic targets represent inclined layering, curved layering, and a high-contrast curved interface drawn from a fault-labelled family. With 2-to-10 Hz continuation, the generative-residual model attains velocity RMSEs of 218.7, 108.0, and 341.9 m/s, compared with 232.5, 100.3, and 412.9 m/s for Grid+TV. Thus the latter wins on curved layering, while the former improves model RMSE by 5.9% and 17.2% on the other targets. Grid+TV nevertheless predicts held-out 10 Hz shots better on all three. SIREN is worse than the common background in every cell; continuation worsens two targets but improves one. A five-weight TV sensitivity study preserves the CurveFault model-error ranking. An equal-budget variational extension fails both recovery and interval coverage. These results establish conditional rankings, not universal superiority or an isolated spectral-bias mechanism. Full-grid residuals also prevent attributing gains uniquely to the generator. We advocate joint reporting of model recovery, unseen-shot prediction, optimization cost, and negative outcomes when auditing neural FWI.
ER -