Posterior consistency for Gaussian process approximations of Bayesian posterior distributions

DOI10.1090/MCOM/3244zbMATH Open1429.60040arXiv1603.02004OpenAlexW2963242019MaRDI QIDQ4600705FDOQ4600705

Authors: A. L. Teckentrup, A. M. Stuart 

Publication date: 12 January 2018

Published in: Mathematics of Computation (Search for Journal in Brave)

Abstract: We study the use of Gaussian process emulators to approximate the parameter-to-observation map or the negative log-likelihood in Bayesian inverse problems. We prove error bounds on the Hellinger distance between the true posterior distribution and various approximations based on the Gaussian process emulator. Our analysis includes approximations based on the mean of the predictive process, as well as approximations based on the full Gaussian process emulator. Our results show that the Hellinger distance between the true posterior and its approximations can be bounded by moments of the error in the emulator. Numerical results confirm our theoretical findings.

Full work available at URL: https://arxiv.org/abs/1603.02004

Recommendations

• Posterior consistency of Gaussian process prior for nonparametric binary regression
• Posterior consistency of logistic Gaussian process priors in density estimation
• Propriety of the reference posterior distribution in Gaussian process modeling
• Gaussian approximation of general non-parametric posterior distributions
• Rates of contraction of posterior distributions based on Gaussian process priors
• Approximate maximum a posteriori with Gaussian process priors
• Lower bounds for posterior rates with Gaussian process priors
• On the stability of Bayes estimators for Gaussian processes
• Posterior consistency for some semi-parametric problems

zbMATH Keywords

Gaussian process regressioninverse problemposterior consistencysurrogate modelBayesian approach

Mathematics Subject Classification ID

Nonparametric regression and quantile regression (62G08) Gaussian processes (60G15) Numerical integration (65D30) Numerical interpolation (65D05) Numerical solution to inverse problems in abstract spaces (65J22)

Cites Work

• Interpolation of spatial data. Some theory for kriging
• Design and analysis of computer experiments. With comments and a rejoinder by the authors
• Bayesian calibration of computer models. (With discussion)
• Gaussian processes for machine learning.
• The Geometry of Random Fields
• Theory of Reproducing Kernels
• Spatial variation. 2nd ed
• The pseudo-marginal approach for efficient Monte Carlo computations
• Statistical and computational inverse problems.
• Title not available (Why is that?)
• Title not available (Why is that?)
• Monte Carlo sampling methods using Markov chains and their applications
• Can local particle filters beat the curse of dimensionality?
• Title not available (Why is that?)
• Title not available (Why is that?)
• Equation of State Calculations by Fast Computing Machines
• Riemann Manifold Langevin and Hamiltonian Monte Carlo Methods
• Scattered Data Approximation
• Combining Field Data and Computer Simulations for Calibration and Prediction
• Inverse problems: a Bayesian perspective
• Approximation errors and model reduction with an application in optical diffusion tomography
• Stochastic Equations in Infinite Dimensions
• Title not available (Why is that?)
• Analytic regularity and polynomial approximation of parametric and stochastic elliptic PDE's
• Sparse adaptive approximation of high dimensional parametric initial value problems
• A Stochastic Collocation Method for Elliptic Partial Differential Equations with Random Input Data
• Modeling uncertainty in flow simulations via generalized polynomial chaos.
• Interpolation of spatial data -- a stochastic or a deterministic problem?
• MCMC methods for functions: modifying old algorithms to make them faster
• Title not available (Why is that?)
• Model Reduction for Large-Scale Systems with High-Dimensional Parametric Input Space
• Local error estimates for radial basis function interpolation of scattered data
• Multi-output separable Gaussian process: towards an efficient, fully Bayesian paradigm for uncertainty quantification
• Stochastic spectral methods for efficient Bayesian solution of inverse problems
• Sobolev bounds on functions with scattered zeros, with applications to radial basis function surface fitting
• Sobolev error estimates and a Bernstein inequality for scattered data interpolation via radial basis functions
• Active Subspaces
• Sparsity in Bayesian inversion of parametric operator equations
• Active subspace methods in theory and practice: applications to kriging surfaces
• Sequential Design of Computer Experiments for the Solution of Bayesian Inverse Problems

Cited In (56)

• A MCMC method based on surrogate model and Gaussian process parameterization for infinite Bayesian PDE inversion
• A probabilistic framework for multidisciplinary design: application to the hydrostructural optimization of supercavitating hydrofoils
• Fast sampling of parameterised Gaussian random fields
• Convergence rates for a class of estimators based on Stein's method
• Randomized approaches to accelerate MCMC algorithms for Bayesian inverse problems
• GParareal: a time-parallel ODE solver using Gaussian process emulation
• A modern retrospective on probabilistic numerics
• Adaptive multi-fidelity polynomial chaos approach to Bayesian inference in inverse problems
• Sequential Design of Computer Experiments for the Solution of Bayesian Inverse Problems
• FINANCIAL TIME SERIES USING NONLINEAR DIFFERENTIAL EQUATION OF GAUSSIAN DISTRIBUTION PROBABILITY DENSITY
• Residual-based error correction for neural operator accelerated Infinite-dimensional Bayesian inverse problems
• Convergence of spectral likelihood approximation based on q-Hermite polynomials for Bayesian inverse problems
• Merging MCMC subposteriors through Gaussian-process approximations
• Numerical Gaussian Processes for Time-Dependent and Nonlinear Partial Differential Equations
• Bayesian Probabilistic Numerical Methods in Time-Dependent State Estimation for Industrial Hydrocyclone Equipment
• Bayesian Parameter Identification in Cahn--Hilliard Models for Biological Growth
• Parallel Gaussian process surrogate Bayesian inference with noisy likelihood evaluations
• Stein variational gradient descent with local approximations
• Markov chain generative adversarial neural networks for solving Bayesian inverse problems in physics applications
• Surrogate modeling for Bayesian inverse problems based on physics-informed neural networks
• Asymptotic Bounds for Smoothness Parameter Estimates in Gaussian Process Interpolation
• Consistency of empirical Bayes and kernel flow for hierarchical parameter estimation
• Sampling-free Bayesian inversion with adaptive hierarchical tensor representations
• Multilevel Adaptive Sparse Leja Approximations for Bayesian Inverse Problems
• Bayesian Inverse Problems Are Usually Well-Posed
• Alternative posterior consistency results in nonparametric binary regression using Gaussian process priors
• An Adaptive Surrogate Modeling Based on Deep Neural Networks for Large-Scale Bayesian Inverse Problems
• Data-driven forward discretizations for Bayesian inversion
• Bayesian inversion using adaptive polynomial chaos kriging within subset simulation
• Title not available (Why is that?)
• Error Bounds for Some Approximate Posterior Measures in Bayesian Inference
• Bayesian Model Calibration with Interpolating Polynomials based on Adaptively Weighted Leja Nodes
• Propriety of the reference posterior distribution in Gaussian process modeling
• Well-Posed Bayesian Inverse Problems: Priors with Exponential Tails
• Rate-optimal refinement strategies for local approximation MCMC
• Consistent online Gaussian process regression without the sample complexity bottleneck
• Adaptive Design and Analysis Via Partitioning Trees for Emulation of a Complex Computer Code
• Chebyshev-Legendre spectral method and inverse problem analysis for the space fractional Benjamin-Bona-Mahony equation
• Sequential Design of Computer Experiments for the Computation of Bayesian Model Evidence
• The SPDE approach to Matérn fields: graph representations
• Adaptive sampling-based quadrature rules for efficient Bayesian prediction
• Calibrate, emulate, sample
• An Acceleration Strategy for Randomize-Then-Optimize Sampling Via Deep Neural Networks
• Random Forward Models and Log-Likelihoods in Bayesian Inverse Problems
• On the local Lipschitz stability of Bayesian inverse problems
• Solving inverse problems using data-driven models
• Deep Learning in High Dimension: Neural Network Expression Rates for Analytic Functions in \(\pmb{L^2(\mathbb{R}^d,\gamma_d)}\)
• Convergence of Gaussian Process Regression with Estimated Hyper-Parameters and Applications in Bayesian Inverse Problems
• Title not available (Why is that?)
• Uniform \(L_{1}\) posterior consistency in compact Gaussian shift experiments
• Gaussian processes for Bayesian inverse problems associated with linear partial differential equations
• Adaptive neural network surrogate model for solving the nonlinear elastic inverse problem via Bayesian inference
• Optimal experimental design: formulations and computations
• Parameter inference based on Gaussian processes informed by nonlinear partial differential equations
• Posterior manifolds over prior parameter regions: beyond pointwise sensitivity assessments for posterior statistics from MCMC inference
• Probabilistic error estimate for numerical discretization of high-index saddle dynamics with inaccurate models

This page was built for publication: Posterior consistency for Gaussian process approximations of Bayesian posterior distributions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4600705)