Self-attention summary networks for subsurface velocity-model building from common-image gathers
Abstract
Common-image gathers (CIGs) contain physically meaningful information about velocity-model errors through reflector focusing and residual moveout, but in conventional imaging workflows they are typically used only as diagnostic tools. In this work, we propose a multiscale self-attention summary network that maps high-dimensional 3D CIG volumes into compact conditioning embeddings for probabilistic subsurface velocity inversion. These learned embeddings preserve offset-dependent kinematic structure and spatial coherence while reducing variability caused by background-velocity mismatch. Conditioned on these summary embeddings, a flow-matching model learns a transport from a Gaussian source distribution to the posterior distribution of plausible velocity fields. Numerical experiments show that, compared with direct conditioning on raw CIGs, the proposed summary network improves posterior velocity inference. In particular, the multiscale attention design provides greater robustness to background-model mismatch, yielding more accurate posterior reconstructions and lower predictive uncertainty.
Related
- Simulation of Groundwater Pollutant Transport Characteristics in Mining Areas Driven by Subsurface Flow Velocity Fields.
- Shallow-to-deep velocity model building via diffusion models-Part II: Realistic scenarios
- Shallow-to-deep velocity model building via diffusion models-Part I: Method and Proof of concept
- Depth-progressive 3D seismic velocity model building via 2D generative diffusion models
- HypoNet Nankai PS: rapid hypocenter determination for the Nankai Trough using physics-informed neural networks with three-dimensional P- and S-wave velocity models
BibTeX
@misc{zeng2026selfattention,
title = {{Self-attention summary networks for subsurface velocity-model building from common-image gathers}},
author = {Shiqin Zeng and Yunlin Zeng and Abhinav Prakash Gahlot and Zijun Deng and Felix J. Herrmann},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.09282},
eprint = {2610.09282},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.09282},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Self-attention summary networks for subsurface velocity-model building from common-image gathers
AU - Shiqin Zeng
AU - Yunlin Zeng
AU - Abhinav Prakash Gahlot
AU - Zijun Deng
AU - Felix J. Herrmann
PY - 2026
DA - 2026/10/07
PB - arXiv
DO - 10.48550/arxiv.2610.09282
UR - https://arxiv.org/abs/2610.09282
AB - Common-image gathers (CIGs) contain physically meaningful information about velocity-model errors through reflector focusing and residual moveout, but in conventional imaging workflows they are typically used only as diagnostic tools. In this work, we propose a multiscale self-attention summary network that maps high-dimensional 3D CIG volumes into compact conditioning embeddings for probabilistic subsurface velocity inversion. These learned embeddings preserve offset-dependent kinematic structure and spatial coherence while reducing variability caused by background-velocity mismatch. Conditioned on these summary embeddings, a flow-matching model learns a transport from a Gaussian source distribution to the posterior distribution of plausible velocity fields. Numerical experiments show that, compared with direct conditioning on raw CIGs, the proposed summary network improves posterior velocity inference. In particular, the multiscale attention design provides greater robustness to background-model mismatch, yielding more accurate posterior reconstructions and lower predictive uncertainty.
ER -