On the brittleness of Bayesian inference
From MaRDI portal
Abstract: With the advent of high-performance computing, Bayesian methods are increasingly popular tools for the quantification of uncertainty throughout science and industry. Since these methods impact the making of sometimes critical decisions in increasingly complicated contexts, the sensitivity of their posterior conclusions with respect to the underlying models and prior beliefs is a pressing question for which there currently exist positive and negative results. We report new results suggesting that, although Bayesian methods are robust when the number of possible outcomes is finite or when only a finite number of marginals of the data-generating distribution are unknown, they could be generically brittle when applied to continuous systems (and their discretizations) with finite information on the data-generating distribution. If closeness is defined in terms of the total variation metric or the matching of a finite system of generalized moments, then (1) two practitioners who use arbitrarily close models and observe the same (possibly arbitrarily large amount of) data may reach opposite conclusions; and (2) any given prior and model can be slightly perturbed to achieve any desired posterior conclusions. The mechanism causing brittlenss/robustness suggests that learning and robustness are antagonistic requirements and raises the question of a missing stability condition for using Bayesian Inference in a continuous world under finite information.
Recommendations
Cites work
- A General Qualitative Definition of Robustness
- A Necessary and Sufficient Condition for the Existence of Consistent Estimates
- An overview of robust Bayesian analysis. (With discussion)
- Analyse 1: Utilisation des vocalisations pour différencier trois hardes de phoques
- Bayes theorem in the 21st century
- Borrowing strength: Theory powering applications -- a Festschrift for Lawrence D. Brown. Selected papers based on the presentations at the conference, Philadelphia, PA, USA, December 15--17, 2010.
- Brittleness of Bayesian inference and new Selberg formulas
- Brittleness of Bayesian inference under finite information in a continuous world
- Dynamics of Bayesian updating with dependent data and misspecified models
- scientific article; zbMATH DE number 3854205 (Why is no real title available?)
- scientific article; zbMATH DE number 49698 (Why is no real title available?)
- scientific article; zbMATH DE number 1285154 (Why is no real title available?)
- scientific article; zbMATH DE number 1354815 (Why is no real title available?)
- scientific article; zbMATH DE number 947414 (Why is no real title available?)
- scientific article; zbMATH DE number 3994822 (Why is no real title available?)
- scientific article; zbMATH DE number 1911984 (Why is no real title available?)
- scientific article; zbMATH DE number 3214063 (Why is no real title available?)
- scientific article; zbMATH DE number 3246773 (Why is no real title available?)
- scientific article; zbMATH DE number 3061365 (Why is no real title available?)
- scientific article; zbMATH DE number 3063387 (Why is no real title available?)
- Inverse problems: a Bayesian perspective
- Limiting Behavior of Posterior Distributions when the Model is Incorrect
- Linearization of Bayesian robustness problems
- Local sensitivity diagnostics for Bayesian inference
- Maximum Likelihood Estimation of Misspecified Models
- Merging of Opinions with Increasing Information
- Misspecification in infinite-dimensional Bayesian statistics
- New axioms for rigorous Bayesian probability
- NON-NORMALITY AND TESTS ON VARIANCES
- Nonparametric Bernstein-von Mises theorems in Gaussian white noise
- Nonparametrics and robustness in modern statistical inference and time series analysis: A Festschrift in honor of Professor Jana Jurečková
- On Bayes procedures
- On the Asymptotic Behavior of Bayes' Estimates in the Discrete Case
- On the Bernstein-von Mises phenomenon for nonparametric Bayes procedures
- On the Bernstein-von Mises phenomenon in the Gaussian white noise model
- On the Bernstein-von Mises theorem with infinite-dimensional parameters
- On the consistency of Bayes estimates
- One-sided inference about functionals of a density
- Optimal uncertainty quantification
- Prior envelopes based on belief functions
- Probability Theory
- Probability, Frequency and Reasonable Expectation
- Qualitative robustness in abstract inference
- Qualitative robustness in Bayesian inference
- Rates of contraction of posterior distributions based on Gaussian process priors
- Robust Statistics
- Scaled Boolean algebras
- Statistical decision theory and Bayesian analysis. 2nd ed
- The Bernstein-von Mises theorem under misspecification
- The dilation phenomenon in robust Bayesian inference. (With discussion)
- The Nonexistence of Certain Statistical Procedures in Nonparametric Problems
Cited in
(27)- Learning dynamical systems from data: a simple cross-validation perspective. I: Parametric kernel flows
- Uncertainty quantification of the 4th kind; optimal posterior accuracy-uncertainty tradeoff with the minimum enclosing ball
- Do ideas have shape? Idea registration as the continuous limit of artificial neural networks
- Probabilistic learning inference of boundary value problem with uncertainties based on Kullback-Leibler divergence under implicit constraints
- Hyperparameter estimation in Bayesian MAP estimation: parameterizations and consistency
- Data-driven uncertainty quantification for predictive flow and transport modeling using support vector machines
- Rejoinder: ``Probabilistic integration: a role in statistical computation?
- Brittleness of Bayesian inference under finite information in a continuous world
- Model-free data-driven inference in computational mechanics
- Multigrid with Rough Coefficients and Multiresolution Operator Decomposition from Hierarchical Information Games
- Uncertainty quantification in graph-based classification of high dimensional data
- Qualitative robustness in Bayesian inference
- On the local Lipschitz stability of Bayesian inverse problems
- GAN-based priors for quantifying uncertainty in supervised learning
- Bayesian Parameter Identification in Cahn--Hilliard Models for Biological Growth
- Robust Bayesian inference via coarsening
- Bayesian Probabilistic Numerical Methods
- Bayesian numerical homogenization
- Probabilistic-learning-based stochastic surrogate model from small incomplete datasets for nonlinear dynamical systems
- Probabilistic learning constrained by realizations using a weak formulation of Fourier transform of probability measures
- Polynomial-chaos-based conditional statistics for probabilistic learning with heterogeneous data applied to atomic collisions of helium on graphite substrate
- Transient anisotropic kernel for probabilistic learning on manifolds
- Gamblets for opening the complexity-bottleneck of implicit schemes for hyperbolic and parabolic ODEs/PDEs with rough coefficients
- The mathematics of adversarial attacks in AI -- why deep learning is unstable despite the existence of stable neural networks
- Generalized median of means principle for Bayesian inference
- Clusterization in D-optimal designs: the case against linearization
- Brittleness of Bayesian inference and new Selberg formulas
This page was built for publication: On the brittleness of Bayesian inference
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2808261)
