Stability of Gibbs posteriors from the Wasserstein loss for Bayesian full waveform inversion

DOI10.1137/20M1334218zbMATH Open1473.62018arXiv2004.03730OpenAlexW3208551383MaRDI QIDQ5158929FDOQ5158929

Authors: M. M. Dunlop, Yunan Yang 

Publication date: 26 October 2021

Published in: SIAM/ASA Journal on Uncertainty Quantification (Search for Journal in Brave)

Abstract: Recently, the Wasserstein loss function has been proven to be effective when applied to deterministic full-waveform inversion (FWI) problems. We consider the application of this loss function in Bayesian FWI so that the uncertainty can be captured in the solution. Other loss functions that are commonly used in practice are also considered for comparison. Existence and stability of the resulting Gibbs posteriors are shown on function space under weak assumptions on the prior and model. In particular, the distribution arising from the Wasserstein loss is shown to be quite stable with respect to high-frequency noise in the data. We then illustrate the difference between the resulting distributions numerically, using Laplace approximations to estimate the unknown velocity field and uncertainty associated with the estimates.

Full work available at URL: https://arxiv.org/abs/2004.03730

Recommendations

• Optimal transport for seismic full waveform inversion
• Misfit function for full waveform inversion based on the Wasserstein metric with dynamic formulation
• Seismic inversion and the data normalization for optimal transport
• Analysis of optimal transport related misfit functions in seismic imaging

zbMATH Keywords

Bayesian inferenceWasserstein metricprior distributionseismic inversionGibbs posterior

Mathematics Subject Classification ID

Bayesian problems; characterization of Bayes procedures (62C10) Inverse problems in geophysics (86A22) Sensitivity, stability, well-posedness (49K40) Numerical solution to inverse problems in abstract spaces (65J22)

Cites Work

• Title not available (Why is that?)
• Bayesian inference with optimal maps
• Generalization of an inequality by Talagrand and links with the logarithmic Sobolev inequality
• Inverse Problem Theory and Methods for Model Parameter Estimation
• A new class of transport distances between measures
• A Bayesian level set method for geometric inverse problems
• Absorbing boundary conditions for numerical simulation of waves
• Optimal transport for seismic full waveform inversion
• Application of the Wasserstein metric to seismic signals
• On the properties of variational approximations of Gibbs posteriors
• On the well-posedness of Bayesian inverse problems
• Well-posed Bayesian inverse problems and heavy-tailed stable quasi-Banach space priors
• Scalable and efficient algorithms for the propagation of uncertainty from data through inference to prediction for large-scale problems, with application to flow of the antarctic ice sheet
• Comparison between \(W_2\) distance and \(\dot{H}^{-1}\) norm, and localization of Wasserstein distance
• An optimal transport approach for seismic tomography: application to 3D full waveform inversion
• Geometric MCMC for infinite-dimensional inverse problems
• Hierarchical Bayesian level set inversion
• Inference for empirical Wasserstein distances on finite spaces
• Uncertainty quantification in graph-based classification of high dimensional data
• On the convergence of the Laplace approximation and noise-level-robustness of Laplace-based Monte Carlo methods for Bayesian inverse problems
• Calibrating general posterior credible regions
• WASSERSTEIN METRIC-DRIVEN BAYESIAN INVERSION WITH APPLICATIONS TO SIGNAL PROCESSING
• The quadratic Wasserstein metric for inverse data matching
• Bayesian inversion in resin transfer molding
• Reconciling Bayesian and perimeter regularization for binary inversion

Cited In (6)

• Generalized Bayes approach to inverse problems with model misspecification
• Optimal Transport for Parameter Identification of Chaotic Dynamics via Invariant Measures
• Norm-dependent convergence and stability of the inverse scattering series for diffuse and scalar waves
• Object based Bayesian full-waveform inversion for shear elastography
• Applications of optimal transportation in the natural sciences. Abstracts from the workshop held February 21--27, 2021 (online meeting)
• Implicit regularization effects of the Sobolev norms in image processing

This page was built for publication: Stability of Gibbs posteriors from the Wasserstein loss for Bayesian full waveform inversion

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5158929)