34
60

Convergence rates for Penalised Least Squares Estimators in PDE-constrained regression problems

Abstract

We consider PDE constrained nonparametric regression problems in which the parameter ff is the unknown coefficient function of a second order elliptic partial differential operator LfL_f, and the unique solution ufu_f of the boundary value problem \[L_fu=g_1\text{ on } \mathcal O, \quad u=g_2 \text{ on }\partial \mathcal O,\] is observed corrupted by additive Gaussian white noise. Here O\mathcal O is a bounded domain in Rd\mathbb R^d with smooth boundary O\partial \mathcal O, and g1,g2g_1, g_2 are given functions defined on O,O\mathcal O, \partial \mathcal O, respectively. Concrete examples include Lfu=Δu2fuL_fu=\Delta u-2fu (Schr\"odinger equation with attenuation potential ff) and Lfu=div(fu)L_fu=\text{div} (f\nabla u) (divergence form equation with conductivity ff). In both cases, the parameter space \[\mathcal F=\{f\in H^\alpha(\mathcal O)| f > 0\}, ~\alpha>0, \] where Hα(O)H^\alpha(\mathcal O) is the usual order α\alpha Sobolev space, induces a set of non-linearly constrained regression functions {uf:fF}\{u_f: f \in \mathcal F\}. We study Tikhonov-type penalised least squares estimators f^\hat f for ff. The penalty functionals are of squared Sobolev-norm type and thus f^\hat f can also be interpreted as a Bayesian `MAP'-estimator corresponding to some Gaussian process prior. We derive rates of convergence of f^\hat f and of uf^u_{\hat f}, to f,uff, u_f, respectively. We prove that the rates obtained are minimax-optimal in prediction loss. Our bounds are derived from a general convergence rate result for non-linear inverse problems whose forward map satisfies a modulus of continuity condition, a result of independent interest that is applicable also to linear inverse problems, illustrated in an example with the Radon transform.

View on arXiv
Comments on this paper

We use cookies and other tracking technologies to improve your browsing experience on our website, to show you personalized content and targeted ads, to analyze our website traffic, and to understand where our visitors are coming from. See our policy.