Adaptive Bayesian estimation using a Gaussian random field with inverse Gamma bandwidth

Abstract

We consider nonparametric Bayesian estimation inference using a rescaled smooth Gaussian field as a prior for a multidimensional function. The rescaling is achieved using a Gamma variable and the procedure can be viewed as choosing an inverse Gamma bandwidth. The procedure is studied from a frequentist perspective in three statistical settings involving replicated observations (density estimation, regression and classification). We prove that the resulting posterior distribution shrinks to the distribution that generates the data at a speed which is minimax-optimal up to a logarithmic factor, whatever the regularity level of the data-generating distribution. Thus the hierachical Bayesian procedure, with a fixed prior, is shown to be fully adaptive.

Published in at http://dx.doi.org/10.1214/08-AOS678 the Annals of Statistics (http://www.imstat.org/aos/) by the Institute of Mathematical Statistics (http://www.imstat.org)

References in corpus (6)

• Rates of contraction of posterior distributions based on Gaussian process priors
• Convergence rates of posterior distributions for noniid observations
• Reproducing kernel Hilbert spaces of Gaussian priors
• Posterior consistency of Gaussian process prior for nonparametric binary regression
• Bayesian inference with rescaled Gaussian process priors
• Convergence rates for posterior distributions and adaptive estimation

Cited by in corpus (64)

• The prior can generally only be understood in the context of the likelihood
• Adaptive Bayesian multivariate density estimation with Dirichlet mixtures
• Nonparametric Bernstein-von Mises theorems in Gaussian white noise
• Rates of convergence for the posterior distributions of mixtures of Betas and adaptive nonparametric estimation of the density
• Bayesian inverse problems with non-conjugate priors
• Rates of contraction for posterior distributions in $\bolds{L^r}$-metrics, $\bolds{1\le r\le\infty}$
• Anisotropic function estimation using multi-bandwidth Gaussian processes
• Minimax-optimal nonparametric regression in high dimensions
• On adaptive posterior concentration rates
• Bayesian optimal adaptive estimation using a sieve prior
• Frequentist coverage and sup-norm convergence rate in Gaussian process regression
• Adaptive nonparametric Bayesian inference using location-scale mixture priors
• A Nonparametric Bayesian Methodology for Regression Discontinuity Designs
• A Divide-and-Conquer Bayesian Approach to Large-Scale Kriging
• Bayes factor consistency
• Semiparametric Bayesian causal inference
• Reliable emulation of complex functionals by active learning with error control
• Bayesian Detection of Image Boundaries
• Bayesian nonparametric dependent model for partially replicated data: the influence of fuel spills on species diversity
• Data-driven priors and their posterior concentration rates
• Nonparametric Bayes modeling of count processes
• Rate exact Bayesian adaptation with modified block priors
• PAC-Bayesian AUC classification and scoring
• Optimality of Poisson processes intensity learning with Gaussian processes
• Dimension adaptability of Gaussian process models with variable selection and projection
• Optimal Bayesian estimation in random covariate design with a rescaled Gaussian process prior
• Comparing and weighting imperfect models using D-probabilities
• A Closer Look at Covering Number Bounds for Gaussian Kernels
• Bayesian fractional posteriors
• Adaptive Gaussian Predictive Process Approximation
• Laplace approximation for logistic Gaussian process density estimation and regression
• Semiparametric Bernstein-von Mises Theorem: Second Order Studies
• Heavy-Tailed Density Estimation
• Non Parametric Hidden Markov Models with Finite State Space: Posterior Concentration Rates
• Adaptive Convergence Rates of a Dirichlet Process Mixture of Multivariate Normals
• Graph Based Gaussian Processes on Restricted Domains
• Adaptive Bayesian density estimation using Pitman-Yor or normalized inverse-Gaussian process kernel mixtures
• Flexible Bayesian Dynamic Modeling of Correlation and Covariance Matrices
• Extrinsic Gaussian processes for regression and classification on manifolds
• Hierarchical Bayesian Level Set Inversion
• Gaussian Processes with Errors in Variables: Theory and Computation
• Sketching in Bayesian High Dimensional Regression With Big Data Using Gaussian Scale Mixture Priors
• Learning Temporal Evolution of Spatial Dependence with Generalized Spatiotemporal Gaussian Process Models
• Comparison theorems on large-margin learning
• Spike and Slab Pólya tree posterior distributions: adaptive inference
• Exponential-wrapped distributions on symmetric spaces
• Gaussian Approximation of General Nonparametric Posterior Distributions
• Approximation in Hilbert spaces of the Gaussian and related analytic kernels
• Spatiotemporal Besov Priors for Bayesian Inverse Problems
• Posterior contraction rates for constrained deep Gaussian processes in density estimation and classication
• Equivalence of Convergence Rates of Posterior Distributions and Bayes Estimators for Functions and Nonparametric Functionals
• Fast Statistical Leverage Score Approximation in Kernel Ridge Regression
• Spatially Adaptive Ensemble Learning with Calibrated Predictive Uncertainty
• Posterior Contraction Rates for Gaussian Cox Processes with Non-identically Distributed Data
• Bayesian Projected Calibration of Computer Models
• Uncertainty quantification for distributed regression
• Bayesian estimation of nonlinear Hawkes process
• Discussion of "Frequentist coverage of adaptive nonparametric Bayesian credible sets"
• Non-asymptotic Bayesian Minimax Adaptation
• Statistical Guarantees for Transformation Based Models with Applications to Implicit Variational Inference
• Wasserstein convergence in Bayesian deconvolution models
• Sample Complexity Result for Multi-category Classifiers of Bounded Variation
• Posterior Convergence of Gaussian and General Stochastic Process Regression Under Possible Misspecifications
• Wasserstein convergence in Bayesian and frequentist deconvolution models