Bernstein - von Mises theorems for statistical inverse problems I: Schr\"odinger equation

TL;DR

This paper establishes Bernstein-von Mises theorems for Bayesian inverse problems involving the Schrödinger equation, showing that the posterior distribution approximates a Gaussian in the small noise limit, enabling optimal statistical inference.

Contribution

It introduces a nonparametric Bayesian approach and proves a Bernstein-von Mises theorem for the inverse Schrödinger problem, linking Bayesian and frequentist inference in this context.

Findings

Posterior distribution approximates a Gaussian in the small noise limit.

The Gaussian measure has a minimal covariance structure.

Bayesian inference becomes asymptotically optimal and valid.

Abstract

The inverse problem of determining the unknown potential $f>0$ in the partial differential equation $$\frac{\Delta}{2} u - fu =0 \text{ on } \mathcal O ~~\text{s.t. } u = g \text { on } \partial \mathcal O,$$ where $\mathcal O$ is a bounded $C^\infty$-domain in $\mathbb R^d$ and $g>0$ is a given function prescribing boundary values, is considered. The data consist of the solution $u$ corrupted by additive Gaussian noise. A nonparametric Bayesian prior for the function $f$ is devised and a Bernstein - von Mises theorem is proved which entails that the posterior distribution given the observations is approximated in a suitable function space by an infinite-dimensional Gaussian measure that has a `minimal' covariance structure in an information-theoretic sense. As a consequence the posterior distribution performs valid and optimal frequentist statistical inference on $f$ in the small noise limit.