Adaptivity in Bayesian inverse finite element problems: Learning and simultaneous control of discretisation and sampling errors

Research output: Contribution to journalJournal articlepeer-review

Standard

Adaptivity in Bayesian inverse finite element problems : Learning and simultaneous control of discretisation and sampling errors. / Kerfriden, Pierre; Kundu, Abhishek; Claus, Susanne.

In: Materials, Vol. 12, No. 4, 642, 2019.

Research output: Contribution to journalJournal articlepeer-review

Harvard

Kerfriden, P, Kundu, A & Claus, S 2019, 'Adaptivity in Bayesian inverse finite element problems: Learning and simultaneous control of discretisation and sampling errors', Materials, vol. 12, no. 4, 642. https://doi.org/10.3390/ma12040642

APA

Kerfriden, P., Kundu, A., & Claus, S. (2019). Adaptivity in Bayesian inverse finite element problems: Learning and simultaneous control of discretisation and sampling errors. Materials, 12(4), [642]. https://doi.org/10.3390/ma12040642

Vancouver

Kerfriden P, Kundu A, Claus S. Adaptivity in Bayesian inverse finite element problems: Learning and simultaneous control of discretisation and sampling errors. Materials. 2019;12(4). 642. https://doi.org/10.3390/ma12040642

Author

Kerfriden, Pierre ; Kundu, Abhishek ; Claus, Susanne. / Adaptivity in Bayesian inverse finite element problems : Learning and simultaneous control of discretisation and sampling errors. In: Materials. 2019 ; Vol. 12, No. 4.

Bibtex

@article{4201c4b8421f4fd2b175e7a9fb0b8d7d,
title = "Adaptivity in Bayesian inverse finite element problems: Learning and simultaneous control of discretisation and sampling errors",
abstract = "The local size of computational grids used in partial differential equation (PDE)-based probabilistic inverse problems can have a tremendous impact on the numerical results. As a consequence, numerical model identification procedures used in structural or material engineering may yield erroneous, mesh-dependent result. In this work, we attempt to connect the field of adaptive methods for deterministic and forward probabilistic finite-element (FE) simulations and the field of FE-based Bayesian inference. In particular, our target setting is that of exact inference, whereby complex posterior distributions are to be sampled using advanced Markov Chain Monte Carlo (MCMC) algorithms. Our proposal is for the mesh refinement to be performed in a goal-oriented manner. We assume that we are interested in a finite subset of quantities of interest (QoI) such as a combination of latent uncertain parameters and/or quantities to be drawn from the posterior predictive distribution. Next, we evaluate the quality of an approximate inversion with respect to these quantities. This is done by running two chains in parallel: (i) the approximate chain and (ii) an enhanced chain whereby the approximate likelihood function is corrected using an efficient deterministic error estimate of the error introduced by the spatial discretisation of the PDE of interest. One particularly interesting feature of the proposed approach is that no user-defined tolerance is required for the quality of the QoIs, as opposed to the deterministic error estimation setting. This is because our trust in the model, and therefore a good measure for our requirement in terms of accuracy, is fully encoded in the prior. We merely need to ensure that the finite element approximation does not impact the posterior distributions of QoIs by a prohibitively large amount. We will also propose a technique to control the error introduced by the MCMC sampler, and demonstrate the validity of the combined mesh and algorithmic quality control strategy.",
keywords = "Bayesian statistics, Data-driven modelling, Error estimation, Finite element inverse problems, Machine learning, MCMC",
author = "Pierre Kerfriden and Abhishek Kundu and Susanne Claus",
year = "2019",
doi = "10.3390/ma12040642",
language = "English",
volume = "12",
journal = "Materials",
issn = "1996-1944",
publisher = "Molecular Diversity Preservation International, MDPI",
number = "4",

}

RIS

TY - JOUR

T1 - Adaptivity in Bayesian inverse finite element problems

T2 - Learning and simultaneous control of discretisation and sampling errors

AU - Kerfriden, Pierre

AU - Kundu, Abhishek

AU - Claus, Susanne

PY - 2019

Y1 - 2019

N2 - The local size of computational grids used in partial differential equation (PDE)-based probabilistic inverse problems can have a tremendous impact on the numerical results. As a consequence, numerical model identification procedures used in structural or material engineering may yield erroneous, mesh-dependent result. In this work, we attempt to connect the field of adaptive methods for deterministic and forward probabilistic finite-element (FE) simulations and the field of FE-based Bayesian inference. In particular, our target setting is that of exact inference, whereby complex posterior distributions are to be sampled using advanced Markov Chain Monte Carlo (MCMC) algorithms. Our proposal is for the mesh refinement to be performed in a goal-oriented manner. We assume that we are interested in a finite subset of quantities of interest (QoI) such as a combination of latent uncertain parameters and/or quantities to be drawn from the posterior predictive distribution. Next, we evaluate the quality of an approximate inversion with respect to these quantities. This is done by running two chains in parallel: (i) the approximate chain and (ii) an enhanced chain whereby the approximate likelihood function is corrected using an efficient deterministic error estimate of the error introduced by the spatial discretisation of the PDE of interest. One particularly interesting feature of the proposed approach is that no user-defined tolerance is required for the quality of the QoIs, as opposed to the deterministic error estimation setting. This is because our trust in the model, and therefore a good measure for our requirement in terms of accuracy, is fully encoded in the prior. We merely need to ensure that the finite element approximation does not impact the posterior distributions of QoIs by a prohibitively large amount. We will also propose a technique to control the error introduced by the MCMC sampler, and demonstrate the validity of the combined mesh and algorithmic quality control strategy.

AB - The local size of computational grids used in partial differential equation (PDE)-based probabilistic inverse problems can have a tremendous impact on the numerical results. As a consequence, numerical model identification procedures used in structural or material engineering may yield erroneous, mesh-dependent result. In this work, we attempt to connect the field of adaptive methods for deterministic and forward probabilistic finite-element (FE) simulations and the field of FE-based Bayesian inference. In particular, our target setting is that of exact inference, whereby complex posterior distributions are to be sampled using advanced Markov Chain Monte Carlo (MCMC) algorithms. Our proposal is for the mesh refinement to be performed in a goal-oriented manner. We assume that we are interested in a finite subset of quantities of interest (QoI) such as a combination of latent uncertain parameters and/or quantities to be drawn from the posterior predictive distribution. Next, we evaluate the quality of an approximate inversion with respect to these quantities. This is done by running two chains in parallel: (i) the approximate chain and (ii) an enhanced chain whereby the approximate likelihood function is corrected using an efficient deterministic error estimate of the error introduced by the spatial discretisation of the PDE of interest. One particularly interesting feature of the proposed approach is that no user-defined tolerance is required for the quality of the QoIs, as opposed to the deterministic error estimation setting. This is because our trust in the model, and therefore a good measure for our requirement in terms of accuracy, is fully encoded in the prior. We merely need to ensure that the finite element approximation does not impact the posterior distributions of QoIs by a prohibitively large amount. We will also propose a technique to control the error introduced by the MCMC sampler, and demonstrate the validity of the combined mesh and algorithmic quality control strategy.

KW - Bayesian statistics

KW - Data-driven modelling

KW - Error estimation

KW - Finite element inverse problems

KW - Machine learning

KW - MCMC

UR - http://www.scopus.com/inward/record.url?scp=85062215291&partnerID=8YFLogxK

U2 - 10.3390/ma12040642

DO - 10.3390/ma12040642

M3 - Journal article

C2 - 30791661

AN - SCOPUS:85062215291

VL - 12

JO - Materials

JF - Materials

SN - 1996-1944

IS - 4

M1 - 642

ER -

ID: 223456370