Stability of doubly-intractable distributions

DOI10.1214/20-ECP341zbMATH Open1448.60009arXiv2004.07310OpenAlexW3083983705MaRDI QIDQ2201541FDOQ2201541

Daniel Rudolf, Björn Sprungk, Michael Habeck

Publication date: 29 September 2020

Published in: Electronic Communications in Probability (Search for Journal in Brave)

Abstract: Doubly-intractable distributions appear naturally as posterior distributions in Bayesian inference frameworks whenever the likelihood contains a normalizing function $Z$. Having two such functions $Z$ and $widetildeZ$ we provide estimates of the total variation and Wasserstein distance of the resulting posterior probability measures. As a consequence this leads to local Lipschitz continuity w.r.t. $Z$. In the more general framework of a random function $widetildeZ$ we derive bounds on the expected total variation and expected Wasserstein distance. The applicability of the estimates is illustrated within the setting of two representative Monte Carlo recovery scenarios.

Full work available at URL: https://arxiv.org/abs/2004.07310

Recommendations

• On Russian roulette estimates for Bayesian inference with doubly-intractable likelihoods
• On the local Lipschitz stability of Bayesian inverse problems
• Stability and infinitesimal robustness of posterior distributions and posterior quantities
• Efficient computational strategies for doubly intractable problems with applications to Bayesian social networks
• An efficient Markov chain Monte Carlo method for distributions with intractable normalising constants

zbMATH Keywords

Wasserstein distancestabilitydoubly-intractable distributions

Mathematics Subject Classification ID

Convergence of probability measures (60B10) Monte Carlo methods (65C05) Bayesian problems; characterization of Bayes procedures (62C10)

Cites Work

• Optimal Transport
• Title not available (Why is that?)
• An efficient Markov chain Monte Carlo method for distributions with intractable normalising constants
• Inverse problems: a Bayesian perspective
• Perturbation theory for Markov chains via Wasserstein distance
• Bayesian computation for statistical models with intractable normalizing constants
• Noisy Monte Carlo: convergence of Markov chains with approximate transition kernels
• Bayesian Inference in the Presence of Intractable Normalizing Functions
• On the Well-posedness of Bayesian Inverse Problems
• Optimal confidence for Monte Carlo integration of smooth functions
• Perturbation bounds for Monte Carlo within metropolis via restricted approximations
• Solvable integration problems and optimal sample size selection
• On the local Lipschitz stability of Bayesian inverse problems
• Random Forward Models and Log-Likelihoods in Bayesian Inverse Problems

Cited In (2)

• On the theoretical properties of the exchange algorithm
• Bayesian Inverse Problems Are Usually Well-Posed

This page was built for publication: Stability of doubly-intractable distributions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2201541)