On the inference of applying Gaussian process modeling to a deterministic function

DOI10.1214/21-EJS1912zbMATH Open1498.62180arXiv2002.01381MaRDI QIDQ2074282FDOQ2074282

Authors: Wenjia Wang 

Publication date: 9 February 2022

Published in: Electronic Journal of Statistics (Search for Journal in Brave)

Abstract: Gaussian process modeling is a standard tool for building emulators for computer experiments, which are usually used to study deterministic functions, for example, a solution to a given system of partial differential equations. This work investigates applying Gaussian process modeling to a deterministic function from prediction and uncertainty quantification perspectives, where the Gaussian process model is misspecified. Specifically, we consider the case where the underlying function is fixed and from a reproducing kernel Hilbert space generated by some kernel function, and the same kernel function is used in the Gaussian process modeling as the correlation function for prediction and uncertainty quantification. While upper bounds and the optimal convergence rate of prediction in the Gaussian process modeling have been extensively studied in the literature, a comprehensive exploration of convergence rates and theoretical study of uncertainty quantification is lacking. We prove that, if one uses maximum likelihood estimation to estimate the variance in Gaussian process modeling, under different choices of the regularization parameter value, the predictor is not optimal and/or the confidence interval is not reliable. In particular, lower bounds of the prediction error under different choices of the regularization parameter value are obtained. The results indicate that, if one directly applies Gaussian process modeling to a fixed function, the reliability of the confidence interval and the optimality of the predictor cannot be achieved at the same time.

Full work available at URL: https://arxiv.org/abs/2002.01381

Recommendations

• Maximum likelihood estimation and uncertainty quantification for Gaussian process approximation of deterministic functions
• Gaussian process modeling with boundary information
• Computer emulation with nonstationary Gaussian processes
• Flexible Correlation Structure for Accurate Prediction and Uncertainty Quantification in Bayesian Gaussian Process Emulation of a Computer Model
• Fixed-domain asymptotics of the maximum likelihood estimator and the Gaussian process approach for deterministic models

zbMATH Keywords

computer experimentsuncertainty quantificationGaussian process modelingMatérn correlation functions

Mathematics Subject Classification ID

Nonparametric regression and quantile regression (62G08) Asymptotic properties of nonparametric inference (62G20) Inference from spatial processes (62M30) Gaussian processes (60G15)

Cites Work

• Interpolation of spatial data. Some theory for kriging
• The design and analysis of computer experiments.
• Estimation and prediction using generalized Wendland covariance functions under fixed domain asymptotics
• On fixed-domain asymptotics and covariance tapering in Gaussian random field models
• Stochastic kriging for simulation metamodeling
• Maximum likelihood estimation of models for residual covariance in spatial regression
• Gaussian processes for machine learning.
• Nonseparable, Stationary Covariance Functions for Space–Time Data
• The role of the range parameter for estimation and prediction in geostatistics
• Inconsistent Estimation and Asymptotically Equal Interpolations in Model-Based Geostatistics
• Cases for the nugget in modeling computer experiments
• A theoretical framework for calibration in computer models: parametrization, estimation and convergence properties
• Optimal global rates of convergence for nonparametric regression
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Multiresolution functional ANOVA for large-scale, many-input computer experiments
• Composite Gaussian process models for emulating expensive functions
• Single nugget kriging
• An efficient surrogate model for emulation and physics extraction of large eddy simulations
• Concentration inequalities and model selection. Ecole d'Eté de Probabilités de Saint-Flour XXXIII -- 2003.
• Convex functions and their applications. A contemporary approach
• On the consistent separation of scale and variance for Gaussian random fields
• On the uniform convergence of empirical norms and inner products, with application to causal inference
• Space and space-time modeling using process convolutions
• Besov Spaces on Domains in ℝ d
• Interpolation of spatial data -- a stochastic or a deterministic problem?
• Fixed-Domain Asymptotics for Spatial Periodograms
• Supremum norm posterior contraction and credible sets for nonparametric multivariate regression
• Title not available (Why is that?)
• Credible sets in the fixed design model with Brownian motion prior
• Cross validation and maximum likelihood estimations of hyper-parameters of Gaussian processes with model misspecification
• Local error estimates for radial basis function interpolation of scattered data
• Asymptotic properties of a maximum likelihood estimator with data from a Gaussian process
• Title not available (Why is that?)
• Maximum likelihood estimation of parameters under a spatial sampling scheme
• Geometric and transformational properties of Lipschitz domains, Semmes-Kenig-Toro domains, and other classes of finite perimeter domains
• Title not available (Why is that?)
• Maximum likelihood estimation and uncertainty quantification for Gaussian process approximation of deterministic functions
• Closed form representations and properties of the generalised Wendland functions
• On frequentist coverage errors of Bayesian credible sets in moderately high dimensions
• Combining experimental data and computer simulations, with an application to flyer plate experiments
• On the efficiency of selection criteria in spline regression
• Maximum Likelihood Estimation for a Smooth Gaussian Random Field Model
• An alternative measure of the reliability of ordinary kriging estimates
• Asymptotic analysis of covariance parameter estimation for Gaussian processes in the misspecified case
• The reproducing kernel Hilbert space structure of the sample paths of a Gaussian process
• Spline smoothing with an estimated order parameter
• Penalized blind kriging in computer experiments
• Convergence types and rates in generic Karhunen-Loève expansions with applications to sample path properties
• On Prediction Properties of Kriging: Uniform Error Bounds and Robustness
• Title not available (Why is that?)
• Uncertainty quantification in multiscale simulation of woven fiber composites

Cited In (10)

• Error bounds and the asymptotic setting in kernel-based approximation
• Title not available (Why is that?)
• A comparative evaluation of stochastic-based inference methods for Gaussian process models
• On estimating the mean function of a Gaussian process
• A Gaussian Process Regression Model for Distribution Inputs
• Solving Fredholm integral equation of the first kind using Gaussian process regression
• Maximum likelihood estimation and uncertainty quantification for Gaussian process approximation of deterministic functions
• The best approximate solution of Fredholm integral equations of the first kind via Gaussian process regression
• Propriety of the reference posterior distribution in Gaussian process modeling
• Identifying the multifractional function of a Gaussian process

Uses Software

• CRAN
• gss
• mlegp

This page was built for publication: On the inference of applying Gaussian process modeling to a deterministic function

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2074282)