Gaussian approximation of general non-parametric posterior distributions

Abstract

In a general class of Bayesian non-parametric models, we prove that the posterior distribution can be asymptotically approximated by a Gaussian process (GP). Our results apply to non-parametric exponential family that contains both Gaussian and non-Gaussian regression and also hold for both efficient (root-$n$) and inefficient (non-root-$n$) estimations. Our general approximation theorem does not rely on posterior conjugacy and can be verified in a class of GP priors that has a smoothing spline interpretation. In particular, the limiting posterior measure becomes prior free under a Bayesian version of ‘under-smoothing’ condition. Finally, we apply our approximation theorem to examine the asymptotic frequentist properties of Bayesian procedures such as credible regions and credible intervals.

Document Details

Document Type
Pub Defense Publication
Publication Date
Dec 22, 2017
Source ID
10.1093/imaiai/iax017

Entities

People

  • Guang Cheng
  • Zuofeng Shang

Organizations

  • Indiana University – Purdue University Indianapolis
  • National Science Foundation
  • Office of Naval Research
  • Purdue University

Tags

Fields of Study

  • Computer science
  • Mathematics

Readers

  • Statistical inference.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Machine Learning Algorithms
  • Space