29
0

Contraction rates and projection subspace estimation with Gaussian process priors in high dimension

Abstract

This work explores the dimension reduction problem for Bayesian nonparametric regression and density estimation. More precisely, we are interested in estimating a functional parameter ff over the unit ball in Rd\mathbb{R}^d, which depends only on a d0d_0-dimensional subspace of Rd\mathbb{R}^d, with d0<dd_0 < d.It is well-known that rescaled Gaussian process priors over the function space achieve smoothness adaptation and posterior contraction with near minimax-optimal rates. Moreover, hierarchical extensions of this approach, equipped with subspace projection, can also adapt to the intrinsic dimension d0d_0 (\cite{Tokdar2011DimensionAdapt}).When the ambient dimension dd does not vary with nn, the minimax rate remains of the order nβ/(2β+d0)n^{-\beta/(2\beta +d_0)}.%When dd does not vary with nn, the order of the minimax rate remains the same regardless of the ambient dimension dd. However, this is up to multiplicative constants that can become prohibitively large when dd grows. The dependences between the contraction rate and the ambient dimension have not been fully explored yet and this work provides a first insight: we let the dimension dd grow with nn and, by combining the arguments of \cite{Tokdar2011DimensionAdapt} and \cite{Jiang2021VariableSelection}, we derive a growth rate for dd that still leads to posterior consistency with minimax rate.The optimality of this growth rate is then discussed.Additionally, we provide a set of assumptions under which consistent estimation of ff leads to a correct estimation of the subspace projection, assuming that d0d_0 is known.

View on arXiv
Comments on this paper