Escaping Geological Generative Prior Distributions in Bayesian Travel Time Tomography
Abstract
Solutions to geophysical inverse problems are generally non-unique, and greatly benefit from additional geological prior information. Generative Neural Networks (GNNs) are commonly used to parametrise such information, typically restricting solutions to a lower-dimensional latent space. The final inversion result is obtained by mapping the latent space parameters to the original higher-dimensional model space using the GNNs. However, solutions then remain confined to a low-dimensional geological generative manifold; if the true Earth model lies outside of this manifold, standard latent space inversion fails. We propose a Bayesian model space inversion method that uses the generative manifold as a guide to find the solutions that lie outside this manifold. We validate our approach in non-linear travel-time tomography experiments on both simple and more geological structures. By comparing our method with model space inversion using uniform prior probabilities and standard latent-space inversion, we demonstrate that the less informative uniform prior fails to constrain the solution, while the latent space approach fails when the true Earth model lies outside of the generative manifold. Conversely, our new model-space inversion method successfully recovers complex off-manifold geological features while maintaining low solution uncertainty.
Versions
Related
- Distributed Proximal Stein Variational Gradient Descent Algorithm for Large-scale Bayesian Inference in Traveltime Tomography
- Bayesian Inversion Beyond the Generator's Range: Reducing Non-physical Constraints from Generative Neural Networks
- An AI-assisted conditioning and geological interpretation workflow for usage in implicit geological modeling
- Learning to Guide Ontology-Driven Geological Interpretation with Graph Neural Networks
- Emirates Exploration Imager (EXI) Reveals Crater and Boulder Distribution on Deimos: Insights Into Its Dynamical and Geological Evolution
BibTeX
@misc{roy2026escaping,
title = {{Escaping Geological Generative Prior Distributions in Bayesian Travel Time Tomography}},
author = {Pratyay Roy and Xuebin Zhao and Andrew Curtis},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.11962},
eprint = {2610.11962},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.11962},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Escaping Geological Generative Prior Distributions in Bayesian Travel Time Tomography
AU - Pratyay Roy
AU - Xuebin Zhao
AU - Andrew Curtis
PY - 2026
DA - 2026/10/08
PB - arXiv
DO - 10.48550/arxiv.2610.11962
UR - https://arxiv.org/abs/2610.11962
AB - Solutions to geophysical inverse problems are generally non-unique, and greatly benefit from additional geological prior information. Generative Neural Networks (GNNs) are commonly used to parametrise such information, typically restricting solutions to a lower-dimensional latent space. The final inversion result is obtained by mapping the latent space parameters to the original higher-dimensional model space using the GNNs. However, solutions then remain confined to a low-dimensional geological generative manifold; if the true Earth model lies outside of this manifold, standard latent space inversion fails. We propose a Bayesian model space inversion method that uses the generative manifold as a guide to find the solutions that lie outside this manifold. We validate our approach in non-linear travel-time tomography experiments on both simple and more geological structures. By comparing our method with model space inversion using uniform prior probabilities and standard latent-space inversion, we demonstrate that the less informative uniform prior fails to constrain the solution, while the latent space approach fails when the true Earth model lies outside of the generative manifold. Conversely, our new model-space inversion method successfully recovers complex off-manifold geological features while maintaining low solution uncertainty.
ER -