Approximate Langevin Monte Carlo with Adaptation for Bayesian Full-Waveform Inversion

Muhammad Izzatullah; T. Van Leeuwen; Daniel Peter

doi:10.3997/2214-4609.202112443

Approximate Langevin Monte Carlo with Adaptation for Bayesian Full-Waveform Inversion

Muhammad Izzatullah, T. Van Leeuwen, Daniel Peter

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

In this work, we present a proof of concept for Bayesian full-waveform inversion (FWI) in 2-D. This is based on approximate Langevin Monte Carlo sampling with a gradient-based adaptation of the posterior distribution. We apply our method to the Marmousi model, and it reliably recovers important aspects of the posterior, including the statistical moments, and 1-D and 2-D marginals. Depending on the variations of seismic velocities, the posterior can be significantly non-Gaussian, which directly suggest that using a Hessian approximation for uncertainty quantification in FWI may not be sufficient.

Original language	English (US)
Title of host publication	82nd EAGE Annual Conference & Exhibition
Publisher	European Association of Geoscientists & Engineers
DOIs	https://doi.org/10.3997/2214-4609.202112443
State	Published - 2021

Bibliographical note

KAUST Repository Item: Exported on 2021-10-05
Acknowledgements: The first author would like to thank Tristan van Leeuwen at Utrecht University for visiting his research lab, which led to this work, and his continuous support. The research visits and the work reported here were supported by funding from King Abdullah University of Science and Technology (KAUST).

Access to Document

10.3997/2214-4609.202112443

Cite this

@inproceedings{945f99211da7440995a0fdac807cfe1f,

title = "Approximate Langevin Monte Carlo with Adaptation for Bayesian Full-Waveform Inversion",

abstract = "In this work, we present a proof of concept for Bayesian full-waveform inversion (FWI) in 2-D. This is based on approximate Langevin Monte Carlo sampling with a gradient-based adaptation of the posterior distribution. We apply our method to the Marmousi model, and it reliably recovers important aspects of the posterior, including the statistical moments, and 1-D and 2-D marginals. Depending on the variations of seismic velocities, the posterior can be significantly non-Gaussian, which directly suggest that using a Hessian approximation for uncertainty quantification in FWI may not be sufficient.",

author = "Muhammad Izzatullah and {Van Leeuwen}, T. and Daniel Peter",

note = "KAUST Repository Item: Exported on 2021-10-05 Acknowledgements: The first author would like to thank Tristan van Leeuwen at Utrecht University for visiting his research lab, which led to this work, and his continuous support. The research visits and the work reported here were supported by funding from King Abdullah University of Science and Technology (KAUST).",

year = "2021",

doi = "10.3997/2214-4609.202112443",

language = "English (US)",

booktitle = "82nd EAGE Annual Conference & Exhibition",

publisher = "European Association of Geoscientists & Engineers",

}

TY - GEN

T1 - Approximate Langevin Monte Carlo with Adaptation for Bayesian Full-Waveform Inversion

AU - Izzatullah, Muhammad

AU - Van Leeuwen, T.

AU - Peter, Daniel

N1 - KAUST Repository Item: Exported on 2021-10-05 Acknowledgements: The first author would like to thank Tristan van Leeuwen at Utrecht University for visiting his research lab, which led to this work, and his continuous support. The research visits and the work reported here were supported by funding from King Abdullah University of Science and Technology (KAUST).

PY - 2021

Y1 - 2021

N2 - In this work, we present a proof of concept for Bayesian full-waveform inversion (FWI) in 2-D. This is based on approximate Langevin Monte Carlo sampling with a gradient-based adaptation of the posterior distribution. We apply our method to the Marmousi model, and it reliably recovers important aspects of the posterior, including the statistical moments, and 1-D and 2-D marginals. Depending on the variations of seismic velocities, the posterior can be significantly non-Gaussian, which directly suggest that using a Hessian approximation for uncertainty quantification in FWI may not be sufficient.

AB - In this work, we present a proof of concept for Bayesian full-waveform inversion (FWI) in 2-D. This is based on approximate Langevin Monte Carlo sampling with a gradient-based adaptation of the posterior distribution. We apply our method to the Marmousi model, and it reliably recovers important aspects of the posterior, including the statistical moments, and 1-D and 2-D marginals. Depending on the variations of seismic velocities, the posterior can be significantly non-Gaussian, which directly suggest that using a Hessian approximation for uncertainty quantification in FWI may not be sufficient.

UR - http://hdl.handle.net/10754/672094

UR - https://www.earthdoc.org/content/papers/10.3997/2214-4609.202112443

U2 - 10.3997/2214-4609.202112443

DO - 10.3997/2214-4609.202112443

M3 - Conference contribution

BT - 82nd EAGE Annual Conference & Exhibition

PB - European Association of Geoscientists & Engineers

ER -

Approximate Langevin Monte Carlo with Adaptation for Bayesian Full-Waveform Inversion

Abstract

Bibliographical note

Access to Document

Other files and links

Fingerprint

Cite this