We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

math.ST

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Mathematics > Statistics Theory

Title: Dimension free non-asymptotic bounds on the accuracy of high dimensional Laplace approximation

Abstract: This note attempts to revisit the classical results on Laplace approximation in a modern non-asymptotic and dimension free form. Such an extension is motivated by applications to high dimensional statistical and optimization problems. The established results provide explicit non-asymptotic bounds on the quality of a Gaussian approximation of the posterior distribution in total variation distance in terms of the so called \emph{effective dimension} \( p_G \). This value is defined as interplay between information contained in the data and in the prior distribution. In the contrary to prominent Bernstein - von Mises results, the impact of the prior is not negligible and it allows to keep the effective dimension small or moderate even if the true parameter dimension is huge or infinite. We also address the issue of using a Gaussian approximation with inexact parameters with the focus on replacing the Maximum a Posteriori (MAP) value by the posterior mean and design the algorithm of Bayesian optimization based on Laplace iterations. The results are specified to the case of nonlinear inverse problem.
Subjects: Statistics Theory (math.ST); Numerical Analysis (math.NA)
MSC classes: 62F15, 60E15, 65C05, 65K10
Cite as: arXiv:2204.11038 [math.ST]
  (or arXiv:2204.11038v3 [math.ST] for this version)

Submission history

From: Vladimir Spokoiny [view email]
[v1] Sat, 23 Apr 2022 10:03:40 GMT (46kb)
[v2] Mon, 23 May 2022 08:35:28 GMT (50kb)
[v3] Thu, 23 Jun 2022 10:08:45 GMT (58kb)

Link back to: arXiv, form interface, contact.