Convergence Rates for Penalized Least Squares Estimators in PDE Constrained Regression Problems

DOI10.1137/18M1236137MaRDI QIDQ4960995zbMATH OpenOpenAlexFDO

Authors Richard Nickl, Sara Van De Geer, Sven Wang

Publication date 24 April 2020

Published in SIAM/ASA Journal on Uncertainty Quantification (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1809.08818

zbMATH Keywords

Bayesian nonparametricsnonlinear inverse problemsstatistical inference for partial differential equations

Mathematics Subject Classification ID

Schrödinger operator, Schrödinger equation (35J10) Asymptotic properties of nonparametric inference (62G20) Numerical methods for inverse problems for boundary value problems involving PDEs (65N21)

Abstract: We consider PDE constrained nonparametric regression problems in which the parameter $f$ is the unknown coefficient function of a second order elliptic partial differential operator $L_f$, and the unique solution $u_f$ of the boundary value problem [L_fu=g_1 ext{ on } mathcal O, quad u=g_2 ext{ on }partial mathcal O,] is observed corrupted by additive Gaussian white noise. Here $mathcalO$ is a bounded domain in $mathbb{R}^d$ with smooth boundary $partialmathcalO$, and $g_1,g_2$ are given functions defined on $mathcalO,partialmathcalO$, respectively. Concrete examples include $L_{f}u=Deltau-2fu$ (Schr"odinger equation with attenuation potential $f$) and $L_{f}u=extdiv(fablau)$ (divergence form equation with conductivity $f$). In both cases, the parameter space [mathcal F={fin H^alpha(mathcal O)| f > 0}, ~alpha>0, ] where $H^{a}lpha(mathcalO)$ is the usual order $alpha$ Sobolev space, induces a set of non-linearly constrained regression functions $u_f:finmathcalF$. We study Tikhonov-type penalised least squares estimators $hatf$ for $f$. The penalty functionals are of squared Sobolev-norm type and thus $hatf$ can also be interpreted as a Bayesian `MAP'-estimator corresponding to some Gaussian process prior. We derive rates of convergence of $hatf$ and of $u_{hatf}$, to $f,u_f$, 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.

Recommendations

• Convergence rate estimates for penalty methods revisited
• On convergence rates for iteratively regularized procedures with linear penalty terms
• scientific article; zbMATH DE number 7339841
• scientific article; zbMATH DE number 766273
• Convergence rates for regularization functionals with polyconvex integrands
• Convergence rates in regularization for nonlinear ill-posed equations with perturbative data
• scientific article; zbMATH DE number 5052446
• Convergence rates for constrained regression splines
• Penalized least squares approximation problems
• Convergence Rates for Regularized Solutions

Cites work

• An Introduction to Banach Space Theory
• An Inverse Problem for the Steady State Diffusion Equation
• Asymptotic equivalence for nonparametric regression with multivariate and random design
• Asymptotic equivalence of nonparametric regression and white noise
• Bayesian inverse problems with Gaussian priors
• Bayesian inverse problems with non-conjugate priors
• Bernstein-von Mises theorems for statistical inverse problems. I: Schrödinger equation
• Consistency and rates of convergence of nonlinear Tikhonov regularization with random noise
• Convergence rates for Tikhonov regularisation of non-linear ill-posed problems
• Convergence Rates of General Regularization Methods for Statistical Inverse Problems and Applications
• Diffusion coefficients estimation for elliptic partial differential equations
• Efficient nonparametric Bayesian inference for X-ray transforms
• Elliptic Partial Differential Equations of Second Order
• From Brownian Motion to Schrödinger’s Equation
• scientific article; zbMATH DE number 3143880 (Why is no real title available?)
• scientific article; zbMATH DE number 3870002 (Why is no real title available?)
• scientific article; zbMATH DE number 5654889 (Why is no real title available?)
• scientific article; zbMATH DE number 4001918 (Why is no real title available?)
• scientific article; zbMATH DE number 3602126 (Why is no real title available?)
• scientific article; zbMATH DE number 1181255 (Why is no real title available?)
• scientific article; zbMATH DE number 936298 (Why is no real title available?)
• scientific article; zbMATH DE number 1446717 (Why is no real title available?)
• scientific article; zbMATH DE number 3360568 (Why is no real title available?)
• Inverse diffusion theory of photoacoustics
• Inverse problems: a Bayesian perspective
• Iterative regularization methods for nonlinear ill-posed problems
• Least squares estimation with complexity penalties
• MAP estimators and their consistency in Bayesian nonparametric inverse problems
• Mathematical foundations of infinite-dimensional statistical models
• Maximum a posteriori probability estimates in infinite-dimensional Bayesian inverse problems
• Modern regularization methods for inverse problems
• Nonlinear Tikhonov regularization in Hilbert scales for inverse boundary value problems with random noise
• Nonparametric Bayesian posterior contraction rates for discretely observed scalar diffusions
• Nonparametric Bayesian posterior contraction rates for scalar diffusions with high-frequency data
• Nonparametric Bernstein-von Mises theorems in Gaussian white noise
• On the injectivity and linearization of the coefficient-to-solution mapping for elliptic boundary value problems
• Optimal a Posteriori Parameter Choice for Tikhonov Regularization for Solving Nonlinear Ill-Posed Problems
• Penalized estimators for non linear inverse problems
• Posterior consistency for Bayesian inverse problems through stability and regression results
• Tikhonov regularization anda posteriorirules for solving nonlinear ill posed problems
• Tikhonov regularization of nonlinear III-posed problems in hilbert scales

Cited in

(37)

• Statistical guarantees for Bayesian uncertainty quantification in nonlinear inverse problems with Gaussian process priors
• Spectral thresholding for the estimation of Markov chain transition operators
• A novel fractional order model of SARS-CoV-2 and Cholera disease with real data
• Hyperparameter estimation in Bayesian MAP estimation: parameterizations and consistency
• Nonparametric statistical inference for drift vector fields of multi-dimensional diffusions
• Bernstein-von Mises theorems and uncertainty quantification for linear inverse problems
• Finite element representations of Gaussian processes: balancing numerical and statistical accuracy
• A Bernstein–von-Mises theorem for the Calderón problem with piecewise constant conductivities
• Consistency of Bayesian inference with Gaussian process priors for a parabolic inverse problem
• An optimal Bayesian estimator for absorption coefficient in diffuse optical tomography
• Consistency of Bayesian inference with Gaussian process priors in an elliptic inverse problem
• Solving inverse problems using data-driven models
• scientific article; zbMATH DE number 2206841 (Why is no real title available?)
• Convergence Rates for Learning Linear Operators from Noisy Data
• Laplace priors and spatial inhomogeneity in Bayesian inverse problems
• A Bayesian approach for consistent reconstruction of inclusions
• Besov-Laplace priors in density estimation: optimal posterior contraction rates and adaptation
• Consistent inference for diffusions from low frequency measurements
• Consistency of the Bayes method for the inverse scattering problem
• On posterior consistency of data assimilation with Gaussian process priors: the 2D-Navier-Stokes equations
• The Bayesian approach to inverse Robin problems
• Elliptic Bayesian inverse problems on metric graphs
• Lower bounds for nonparametric estimation of ordinary differential equations
• Consistency of the Bayes method for the inverse scattering problem with randomly truncated sieve priors
• Convergence rates of non-stationary and deep Gaussian process regression
• Potential identification via Tikhonov-PINNs
• Convergence rates for the maximum a posteriori estimator in PDE-regression models with random design
• On low frequency inference for diffusions without the hot spots conjecture
• Stochastic convergence analysis of the inverse potential problem
• Consistency of Bayesian inference for a subdiffusion equation
• Statistical algorithms for low-frequency diffusion data: a PDE approach
• Criticality measure-based error estimates for infinite dimensional optimization
• Misspecified Bernstein-von Mises theorem for hierarchical models
• Consistency of variational inference for Besov priors in non-linear inverse problems
• A persistent-homology-based Bayesian prior for potential coefficient reconstruction in an elliptic partial differential equation
• Linear methods for nonlinear inverse problems
• On statistical Calderón problems

This page was built for publication: Convergence Rates for Penalized Least Squares Estimators in PDE Constrained Regression Problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4960995)