Deep generator priors for Bayesian seismic inversion

Fang, Zhilong; Fang, Hongjian; Demanet, Laurent

doi:10.1016/bs.agph.2020.07.002

Cited by 12 publications

(4 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Given observation data, the posterior distribution of the uncertain parameters is computed using Bayes' rule [54]. In the following, we briefly go over this technique [55].…”

Section: Bayesian Inferencementioning

confidence: 99%

Curing Effect on Durability of Cement Mortar with GGBS: Experimental and Numerical Study

et al. 2022

View full text Add to dashboard Cite

In this paper, supplementary cementitious materials are used as a substitute for cement to decrease carbon dioxide emissions. A by-product of the iron manufacturing industry, ground granulated blast-furnace slag (GGBS), known to improve some performance characteristics of concrete, is used as an effective cement replacement to manufacture mortar samples. Here, the influence of curing conditions on the durability of samples including various amounts of GGBS is investigated experimentally and numerically. Twelve high-strength Portland cement CEM I 52.5 N samples were prepared, in which 0%, 45%, 60%, and 80% of cement were substituted by GGBS. In addition, three curing conditions (standard, dry, and cold curing) were applied to the samples. Durability aspects were studied through porosity, permeability, and water absorption. Experimental results indicate that samples cured in standard conditions gave the best performance in comparison to other curing conditions. Furthermore, samples incorporating 45% of GGBS have superior durability properties. Permeability and water absorption were improved by 17% and 18%, respectively, compared to the reference sample. Thereafter, data from capillary suction experiments were used to numerically determine the hydraulic properties based on a Bayesian inversion approach, namely the Markov Chain Monte Carlo method. Finally, the developed numerical model accurately estimates the hydraulic characteristics of mortar samples and greatly matches the measured water inflow over time through the samples.

show abstract

“…Given observation data, the posterior distribution of the uncertain parameters is computed using Bayes' rule [54]. In the following, we briefly go over this technique [55].…”

Section: Bayesian Inferencementioning

confidence: 99%

Curing Effect on Durability of Cement Mortar with GGBS: Experimental and Numerical Study

et al. 2022

View full text Add to dashboard Cite

show abstract

“…While these approaches have proven to be useful and are theoretically well understood [73], there is always a risk of a biased outcome something we would like to avoid. On the other hand, using pretrained generative networks as priors while proving to be effective [20,27,35,44,74], their success hinges on the quality of pretraining and having access to a fully representative training data that accurately captures the prior distribution. Since we are dealing with highly complex heterogeneity of the Earth subsurface to which we have limited access, we will stay away from data-driven methods to train a neural network to act as a prior.…”

Section: Probabilistic Imaging With Bayesian Inferencementioning

confidence: 99%

“…While effective in controlled settings, handcrafted priors might introduce unwanted bias to the solution. Recent deep-learning based approaches [20][21][22][23][24][25][26][27][28][29][30][31], on the other hand, learn a prior distribution from available data 1 . While certainly providing a better description of the available prior information when compared to generic handcrafted priors, they may affect the results more seriously when out-of-distribution data is considered, e.g., when the training data is not fully representative of a given scenario.…”

Section: Introductionmentioning

confidence: 99%

Deep Bayesian inference for seismic imaging with tasks

Siahkoohi¹,

Rizzuti²,

Herrmann³

2021

Preprint

View full text Add to dashboard Cite

We propose to use techniques from Bayesian inference and deep neural networks to translate uncertainty in seismic imaging to uncertainty in tasks performed on the image, such as horizon tracking. Seismic imaging is an ill-posed inverse problem because of unavoidable bandwidth and aperture limitations, which that is hampered by the presence of noise and linearization errors. Many regularization methods, such as transform-domain sparsity promotion, have been designed to deal with the adverse effects of these errors, however, these methods run the risk of biasing the solution and do not provide information on uncertainty in the image space and how this uncertainty impacts certain tasks on the image. A systematic approach is proposed to translate uncertainty due to noise in the data to confidence intervals of automatically tracked horizons in the image. The uncertainty is characterized by a convolutional neural network (CNN) and to assess these uncertainties, samples are drawn from the posterior distribution of the CNN weights, used to parameterize the image. Compared to traditional priors, in the literature it is argued that these CNNs introduce a flexible inductive bias that is a surprisingly good fit for many diverse domains in imaging. The method of stochastic gradient Langevin dynamics is employed to sample from the posterior distribution. This method is designed to handle large scale Bayesian inference problems with computationally expensive forward operators as in seismic imaging. Aside from offering a robust alternative to maximum a posteriori estimate that is prone to overfitting, access to these samples allow us to translate uncertainty in the image, due to noise in the data, to uncertainty on the tracked horizons. For instance, it admits estimates for the pointwise standard deviation on the image and for confidence intervals on its automatically tracked horizons.

show abstract

“…In seismic inversion, several groups have experimented with directly mapping data to model using deep learning (Araya-Polo et al, 2018;Yang & Ma, 2019;Wu & Lin, 2019;Zhang & Lin, 2020;Kazei et al, 2020). Within Bayesian seismic inversion framework, deep learning has been applied for formulating priors (Herrmann et al, 2019;Mosser et al, 2020;Fang et al, 2020). Other groups use deep learning as a signal processing step to acquire reasonable data for inversion.…”

Section: Introductionmentioning

confidence: 99%

Deep learning for low frequency extrapolation of multicomponent data in elastic full waveform inversion

Sun,

Demanet

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Full waveform inversion (FWI) strongly depends on an accurate starting model to succeed. This is particularly true in the elastic regime: The cycle-skipping phenomenon is more severe in elastic FWI compared to acoustic FWI, due to the short S-wave wavelength. In this paper, we extend our work on extrapolated FWI (EFWI) by proposing to synthesize the low frequencies of multi-component elastic seismic records, and use those "artificial" low frequencies to seed the frequency sweep of elastic FWI. Our solution involves deep learning: we separately train the same convolutional neural network (CNN) on two training datasets, one with vertical components and one with horizontal components of particle velocities, to extrapolate the low frequencies of elastic data. The architecture of this CNN is designed with a large receptive field, by either large convolutional kernels or dilated convolution. Numerical examples on the Marmousi2 model show that the 2-4Hz low frequency data extrapolated from band-limited data above 4Hz provide good starting models for elastic FWI of P-wave and S-wave velocities. Additionally, we study the generalization ability of the proposed neural network over different physical models. For elastic test data, collecting the training dataset by elastic simulation shows better extrapolation accuracy than acoustic simulation, i.e., a smaller generalization gap.

show abstract

Deep generator priors for Bayesian seismic inversion

Cited by 12 publications

References 29 publications

Curing Effect on Durability of Cement Mortar with GGBS: Experimental and Numerical Study

Curing Effect on Durability of Cement Mortar with GGBS: Experimental and Numerical Study

Deep Bayesian inference for seismic imaging with tasks

Deep learning for low frequency extrapolation of multicomponent data in elastic full waveform inversion

Contact Info

Product

Resources

About