Continuum limits of posteriors in graph Bayesian inverse problems

DOI10.1137/17M1138005zbMATH Open1416.28003arXiv1706.07193OpenAlexW2963266525WikidataQ129511905 ScholiaQ129511905MaRDI QIDQ3176426FDOQ3176426

Nicolás García Trillos, Daniel Sanz-Alonso

Publication date: 20 July 2018

Published in: SIAM Journal on Mathematical Analysis (Search for Journal in Brave)

Abstract: We consider the problem of recovering a function input of a differential equation formulated on an unknown domain $M$. We assume to have access to a discrete domain $M_n=x_1,dots,x_{n}subsetM$, and to noisy measurements of the output solution at $plen$ of those points. We introduce a graph-based Bayesian inverse problem, and show that the graph-posterior measures over functions in $M_n$ converge, in the large $n$ limit, to a posterior over functions in $M$ that solves a Bayesian inverse problem with known domain. The proofs rely on the variational formulation of the Bayesian update, and on a new topology for the study of convergence of measures over functions on point clouds to a measure over functions on the continuum. Our framework, techniques, and results may serve to lay the foundations of robust uncertainty quantification of graph-based tasks in machine learning. The ideas are presented in the concrete setting of recovering the initial condition of the heat equation on an unknown manifold.

Full work available at URL: https://arxiv.org/abs/1706.07193

Recommendations

• Consistency of Bayesian inference with Gaussian process priors in an elliptic inverse problem
• Consistency of Bayesian inference with Gaussian process priors for a parabolic inverse problem
• Posterior consistency for Bayesian inverse problems through stability and regression results
• Bernstein-von Mises theorems for statistical inverse problems. I: Schrödinger equation
• Bernstein-von Mises theorems and uncertainty quantification for linear inverse problems

zbMATH Keywords

machine learningBayesian inverse problemheat equationgraph Laplacian

Mathematics Subject Classification ID

Bayesian inference (62F15) Applications of functional analysis in probability theory and statistics (46N30) Spaces of measures, convergence of measures (28A33)

Cites Work

• Importance sampling: intrinsic dimension and computational cost
• Inverse problems: a Bayesian perspective
• Title not available (Why is that?)
• Gamma-convergence of gradient flows on Hilbert and metric spaces and applications
• Continuum limit of total variation on point clouds
• On the Rate of Convergence of Empirical Measures in ∞-transportation Distance
• On the heat flow on metric measure spaces: existence, uniqueness and stability
• Title not available (Why is that?)
• A graph discretization of the Laplace-Beltrami operator
• Uncertainty Quantification and Weak Approximation of an Elliptic Inverse Problem
• Bayesian inverse problems for functions and applications to fluid mechanics
• Large‐Scale Inverse Problems and Quantification of Uncertainty
• Fatou's Lemma for Weakly Converging Probabilities
• The Bayesian formulation and well-posedness of fractional elliptic inverse problems
• Uncertainty Quantification in Graph-Based Classification of High Dimensional Data

Cited In (20)

• Variational Limits of $k$-NN Graph-Based Functionals on Data Clouds
• Kernel Methods for Bayesian Elliptic Inverse Problems on Manifolds
• Optimization on Manifolds via Graph Gaussian Processes
• Analysis of stochastic gradient descent in continuous time
• Asymptotic spectra of large (grid) graphs with a uniform local structure. I: Theory
• Data-driven forward discretizations for Bayesian inversion
• Title not available (Why is that?)
• Title not available (Why is that?)
• Graph-based prior and forward models for inverse problems on manifolds with boundaries
• Lipschitz Regularity of Graph Laplacians on Random Data Clouds
• Fatou's Lemma for Weakly Converging Measures under the Uniform Integrability Condition
• Convergence theorems for varying measures under convexity conditions and applications
• Rates of convergence for Laplacian semi-supervised learning with low labeling rates
• The SPDE approach to Matérn fields: graph representations
• The Bayesian update: variational formulations and gradient flows
• Analysis of $p$-Laplacian Regularization in Semisupervised Learning
• A Maximum Principle Argument for the Uniform Convergence of Graph Laplacian Regressors
• Convergence for varying measures
• Analysis of a Computational Framework for Bayesian Inverse Problems: Ensemble Kalman Updates and MAP Estimators under Mesh Refinement
• Title not available (Why is that?)

This page was built for publication: Continuum limits of posteriors in graph Bayesian inverse problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3176426)