On the ergodicity of the adaptive Metropolis algorithm on unbounded domains

DOI10.1214/10-AAP682MaRDI QIDQ614121zbMATH OpenOpenAlexWikidataFDO

Authors Eero Saksman, Matti Vihola

Publication date 27 December 2010

Published in The Annals of Applied Probability (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/0806.2933

zbMATH Keywords

adaptive Markov chain Monte Carlo methodconvergenceergodicityMetropolis algorithmstochastic approximation

Mathematics Subject Classification ID

Computational methods in Markov chains (60J22) Monte Carlo methods (65C05) Numerical analysis or methods applied to Markov chains (65C40) Stochastic stability in control theory (93E15) Stochastic learning and adaptive control (93E35)

Abstract: This paper describes sufficient conditions to ensure the correct ergodicity of the Adaptive Metropolis (AM) algorithm of Haario, Saksman and Tamminen [Bernoulli 7 (2001) 223--242] for target distributions with a noncompact support. The conditions ensuring a strong law of large numbers require that the tails of the target density decay super-exponentially and have regular contours. The result is based on the ergodicity of an auxiliary process that is sequentially constrained to feasible adaptation sets, independent estimates of the growth rate of the AM chain and the corresponding geometric drift constants. The ergodicity result of the constrained process is obtained through a modification of the approach due to Andrieu and Moulines [Ann. Appl. Probab. 16 (2006) 1462--1505].

Review text

The authors describe sufficient conditions to ensure the correct ergodicity of the adaptive Metropolis algorithm for target distributions with a noncompact support; see \textit{H.~Haario}, \textit{E.~Saksman}, and \textit{J.~Tamminen} [Bernoulli 7, No.~2, 223--242 (2001; Zbl 0989.65004)]; see also \textit{C.~Andrieu} and \textit{É.~Moulines} [Ann. Appl. Probab. 16, No.~3, 1462--1505 (2006; Zbl 1114.65001)]. The work is essentially self-contained, and assumes little background knowledge.

Recommendations

• On the stability and ergodicity of adaptive scaling Metropolis algorithms
• On the containment condition for adaptive Markov chain Monte Carlo algorithms
• Can the adaptive Metropolis algorithm collapse without the covariance lower bound?
• Limit theorems for some adaptive MCMC algorithms with subgeometric kernels
• On the efficiency of adaptive MCMC algorithms

Cites work

• An adaptive Metropolis algorithm
• Computable bounds for geometric convergence rates of Markov chains
• Coupling and Ergodicity of Adaptive Markov Chain Monte Carlo Algorithms
• Geometric convergence and central limit theorems for multidimensional Hastings and Metropolis algorithms
• Geometric ergodicity of Metropolis algorithms
• Markov Chain Monte Carlo Methods for High Dimensional Inversion in Remote Sensing
• On adaptive Markov chain Monte Carlo algorithms
• On the ergodicity of the adaptive Metropolis algorithm on unbounded domains
• On the ergodicity properties of some adaptive MCMC algorithms
• Renewal theory and computable convergence rates for geometrically erdgodic Markov chains
• Some Multivariate Chebyshev Inequalities with Extensions to Continuous Parameter Processes
• Stability of Stochastic Approximation under Verifiable Conditions

Cited in

(33)

• Robust adaptive Metropolis algorithm with coerced acceptance rate
• Adaptive Metropolis algorithm using variational Bayesian adaptive Kalman filter
• Ergodicity of combocontinuous adaptive MCMC algorithms
• Adaptive Gibbs samplers and related MCMC methods
• Conditional particle filters with diffuse initial distributions
• Geometric adaptive Monte Carlo in random environment
• An adaptive multiple-try Metropolis algorithm
• On the stability of some controlled Markov chains and its applications to stochastic approximation with Markovian dynamic
• Stochastic proximal-gradient algorithms for penalized mixed models
• Convergence properties of pseudo-marginal Markov chain Monte Carlo algorithms
• Dimension-independent likelihood-informed MCMC
• Grapham: graphical models with adaptive random walk Metropolis algorithms
• A central limit theorem for adaptive and interacting Markov chains
• Markovian stochastic approximation with expanding projections
• A note on formal constructions of sequential conditional couplings
• Convergence of Markovian stochastic approximation with discontinuous dynamics
• Accelerated dimension-independent adaptive metropolis
• Quantitative convergence rates for subgeometric Markov chains
• On the containment condition for adaptive Markov chain Monte Carlo algorithms
• Limit theorems for some adaptive MCMC algorithms with subgeometric kernels. II
• Convergence of adaptive and interacting Markov chain Monte Carlo algorithms
• Kernel estimators of asymptotic variance for adaptive Markov chain Monte Carlo
• Bayesian computation: a summary of the current state, and samples backwards and forwards
• Limit theorems for some adaptive MCMC algorithms with subgeometric kernels
• On the ergodicity of the adaptive Metropolis algorithm on unbounded domains
• Can the adaptive Metropolis algorithm collapse without the covariance lower bound?
• On the stability and ergodicity of adaptive scaling Metropolis algorithms
• Diffusion approximations and control variates for MCMC
• Accelerating MCMC algorithms
• Almost sure convergence rates of adaptive increasingly rare Markov chain Monte Carlo
• An invitation to adaptive Markov chain Monte Carlo convergence theory
• On adaptive Metropolis-Hastings methods
• Stability of adversarial Markov chains, with an application to adaptive MCMC algorithms

This page was built for publication: On the ergodicity of the adaptive Metropolis algorithm on unbounded domains

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q614121)