Saved in:
Bibliographic Details
Main Authors: Nieman, Dennis, Szabo, Botond, van Zanten, Harry
Format: Preprint
Published: 2021
Subjects:
Online Access:https://arxiv.org/abs/2109.10755
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911399702167552
author Nieman, Dennis
Szabo, Botond
van Zanten, Harry
author_facet Nieman, Dennis
Szabo, Botond
van Zanten, Harry
contents We study the theoretical properties of a variational Bayes method in the Gaussian Process regression model. We consider the inducing variables method introduced by Titsias (2009a) and derive sufficient conditions for obtaining contraction rates for the corresponding variational Bayes (VB) posterior. As examples we show that for three particular covariance kernels (Matérn, squared exponential, random series prior) the VB approach can achieve optimal, minimax contraction rates for a sufficiently large number of appropriately chosen inducing variables. The theoretical findings are demonstrated by numerical experiments.
format Preprint
id arxiv_https___arxiv_org_abs_2109_10755
institution arXiv
publishDate 2021
record_format arxiv
spellingShingle Contraction rates for sparse variational approximations in Gaussian process regression
Nieman, Dennis
Szabo, Botond
van Zanten, Harry
Statistics Theory
We study the theoretical properties of a variational Bayes method in the Gaussian Process regression model. We consider the inducing variables method introduced by Titsias (2009a) and derive sufficient conditions for obtaining contraction rates for the corresponding variational Bayes (VB) posterior. As examples we show that for three particular covariance kernels (Matérn, squared exponential, random series prior) the VB approach can achieve optimal, minimax contraction rates for a sufficiently large number of appropriately chosen inducing variables. The theoretical findings are demonstrated by numerical experiments.
title Contraction rates for sparse variational approximations in Gaussian process regression
topic Statistics Theory
url https://arxiv.org/abs/2109.10755