aiqbench analysis
This research demonstrates that integrating neural operators as preconditioners can significantly improve Full-Waveform Inversion (FWI) accuracy, provided the model accounts for objective-space consistency rather than simple image prediction. The study highlights a critical failure mode in standard Fourier Neural Operators when applied to this physics-informed task, suggesting that Jacobian-informed architectures are necessary for stable gradient updates. These findings offer actionable insights for advancing hybrid AI-physics workflows in geophysical imaging and inverse problems.
Full Waveform Inversion with a Learned Preconditioner
arXiv:2610.15331v1 Announce Type: new Abstract: Full-waveform inversion (FWI) is a powerful approach for reconstructing subsurface velocity models, but its practical performance is limited by ill-conditioning, sensitivity to the update direction, and the high cost of curvature-based methods. In this project, we investigate whether a neural operator trained on reverse-time-migration (RTM) images can provide a useful learned preconditioner for FWI. We compare projected FWI against neural-preconditioned FWI using either a standard Fourier neural operator (FNO) or a Jacobian-informed FNO (JI-FNO). The neural preconditioner applies the normal operator of the learned RTM map to modify the true Devito FWI gradient. Our experiments show that the JI-FNO preconditioner achieves the lowest model error while maintaining data-residual decay comparable to baseline projected FWI. In contrast, the standard FNO preconditioner is unstable and gives worse model recovery, despite using the same neural-preconditioning framework. We also identify a failure mode of a nonlinear neural Gauss Newton-type inner solve: directly matching the neural RTM output to the FWI gradient creates an objective-space mismatch and does not yield a reliable descent direction. These results suggest that forward RTM-image prediction alone is insufficient for learned preconditioning; accurate learned derivatives and objective-consistent use of the neural operator are critical for producing useful inversion updates.
Abstract:Full-waveform inversion (FWI) is a powerful approach for reconstructing subsurface velocity models, but its practical performance is limited by ill-conditioning, sensitivity to the update direction, and the high cost of curvature-based methods. In this project, we investigate whether a neural operator trained on reverse-time-migration (RTM) images can provide a useful learned preconditioner for FWI. We compare projected FWI against neural-preconditioned FWI using either a standard Fourier neural operator (FNO) or a Jacobian-informed FNO (JI-FNO). The neural preconditioner applies the normal operator of the learned RTM map to modify the true Devito FWI gradient. Our experiments show that the JI-FNO preconditioner achieves the lowest model error while maintaining data-residual decay comparable to baseline projected FWI. In contrast, the standard FNO preconditioner is unstable and gives worse model recovery, despite using the same neural-preconditioning framework. We also identify a failure mode of a nonlinear neural Gauss Newton-type inner solve: directly matching the neural RTM output to the FWI gradient creates an objective-space mismatch and does not yield a reliable descent direction. These results suggest that forward RTM-image prediction alone is insufficient for learned preconditioning; accurate learned derivatives and objective-consistent use of the neural operator are critical for producing useful inversion updates.
Machine Learning (cs.LG); Geophysics (physics.geo-ph)
arXiv:2610.15331 [cs.LG]
(or arXiv:2610.15331v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.15331 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 295 words
Related coverage
SPAMoE: Spectrum-Aware Hybrid Operator Framework for Full-Waveform Inversion
arXiv · Jun 9, 2026
ROMNet: a hybrid reduced order modeling and machine learning approach to waveform inversion
arXiv · Aug 27, 2026
Accelerating physics-informed neural networks for full waveform inversion using a hybrid quantum-classical finite-basis architecture
arXiv · Jun 16, 2026
Physical-State-Guided Diffusion Sampling for Full-Waveform Inversion
arXiv · Sep 14, 2026