Spectral gaps for a Metropolis-Hastings algorithm in infinite dimensions (Q473164): Difference between revisions
From MaRDI portal
Created a new Item |
ReferenceBot (talk | contribs) Changed an Item |
||
(7 intermediate revisions by 5 users not shown) | |||
Property / review text | |||
It is well known that the most widely used method for general target measures are Markov chain Monte Carlo (MCMC) algorithms. The main aim of this article is to study the complexity of certain sampling algorithms in high dimensions, more precisely, the Metropolis-Hastings algorithm in infinite dimensions. The computational cost of such an algorithm is the number of necessary steps \(\times\) cost of a step. While for most algorithms the cost of one step grows with the dimension, a major result of this paper is an lgorithm which, when applied to measures, defines via a finite-dimensional approximation of a measure defined by a density with respect to a Gaussian random field, requires a number of steps independent of the dimension in order to achieve a given level of accuracy. The authors consider Metropolis-Hastings MCMC methods and focus on cases when the proposal is either a Gaussian random walk (RWM) for which the covariance equals to that of a reference measure or an Orstain-Uhlenbeck proposal (pCN) for which the reference measure is invariant. Among the main results belong: {\parindent=6mm \begin{itemize}\item[{\(\bullet\)}] Description of \(RWM\) and \(pCN\) algorithms and study of their properties, both positive and negative ones. \item[{\(\bullet\)}] Statement of the weak Harris theorem and a discussion of the relationship between exponential convergence in a Wasserstein distance and \(L^{2}_{\mu}\)-spectral gaps. \item[{\(\bullet\)}] Formal confirmation of a conjecture that \(pCN\) has a convergence rate that is independent of the dimension while \(RWM\) has undesirable dimension dependent properties. \end{itemize}} The results are presented for separable Hilbert spaces, but in fact they do also hold on an arbitrary Banach space. | |||
Property / review text: It is well known that the most widely used method for general target measures are Markov chain Monte Carlo (MCMC) algorithms. The main aim of this article is to study the complexity of certain sampling algorithms in high dimensions, more precisely, the Metropolis-Hastings algorithm in infinite dimensions. The computational cost of such an algorithm is the number of necessary steps \(\times\) cost of a step. While for most algorithms the cost of one step grows with the dimension, a major result of this paper is an lgorithm which, when applied to measures, defines via a finite-dimensional approximation of a measure defined by a density with respect to a Gaussian random field, requires a number of steps independent of the dimension in order to achieve a given level of accuracy. The authors consider Metropolis-Hastings MCMC methods and focus on cases when the proposal is either a Gaussian random walk (RWM) for which the covariance equals to that of a reference measure or an Orstain-Uhlenbeck proposal (pCN) for which the reference measure is invariant. Among the main results belong: {\parindent=6mm \begin{itemize}\item[{\(\bullet\)}] Description of \(RWM\) and \(pCN\) algorithms and study of their properties, both positive and negative ones. \item[{\(\bullet\)}] Statement of the weak Harris theorem and a discussion of the relationship between exponential convergence in a Wasserstein distance and \(L^{2}_{\mu}\)-spectral gaps. \item[{\(\bullet\)}] Formal confirmation of a conjecture that \(pCN\) has a convergence rate that is independent of the dimension while \(RWM\) has undesirable dimension dependent properties. \end{itemize}} The results are presented for separable Hilbert spaces, but in fact they do also hold on an arbitrary Banach space. / rank | |||
Normal rank | |||
Property / Mathematics Subject Classification ID | |||
Property / Mathematics Subject Classification ID: 65C05 / rank | |||
Normal rank | |||
Property / Mathematics Subject Classification ID | |||
Property / Mathematics Subject Classification ID: 65C40 / rank | |||
Normal rank | |||
Property / Mathematics Subject Classification ID | |||
Property / Mathematics Subject Classification ID: 60J22 / rank | |||
Normal rank | |||
Property / Mathematics Subject Classification ID | |||
Property / Mathematics Subject Classification ID: 60G50 / rank | |||
Normal rank | |||
Property / Mathematics Subject Classification ID | |||
Property / Mathematics Subject Classification ID: 60G60 / rank | |||
Normal rank | |||
Property / zbMATH DE Number | |||
Property / zbMATH DE Number: 6371855 / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
Wasserstein spectral gaps | |||
Property / zbMATH Keywords: Wasserstein spectral gaps / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
\(L^{2}\)-spectral gaps | |||
Property / zbMATH Keywords: \(L^{2}\)-spectral gaps / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
Markov chain Monte Carlo in infinite dimensions | |||
Property / zbMATH Keywords: Markov chain Monte Carlo in infinite dimensions / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
weak Harris theorem | |||
Property / zbMATH Keywords: weak Harris theorem / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
random walk | |||
Property / zbMATH Keywords: random walk / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
Gaussian random field | |||
Property / zbMATH Keywords: Gaussian random field / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
sampling algorithm | |||
Property / zbMATH Keywords: sampling algorithm / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
Metropolis-Hastings algorithm | |||
Property / zbMATH Keywords: Metropolis-Hastings algorithm / rank | |||
Normal rank | |||
Property / Wikidata QID | |||
Property / Wikidata QID: Q56894224 / rank | |||
Normal rank | |||
Property / reviewed by | |||
Property / reviewed by: Jaromír Antoch / rank | |||
Normal rank | |||
Property / MaRDI profile type | |||
Property / MaRDI profile type: MaRDI publication profile / rank | |||
Normal rank | |||
Property / arXiv ID | |||
Property / arXiv ID: 1112.1392 / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Q4001807 / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: A New Approach to the Limit Theory of Recurrent Markov Chains / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Q5186515 / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Advanced MCMC methods for sampling on diffusion pathspace / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Optimal scalings for local Metropolis-Hastings chains on nonproduct targets in high dimensions / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: MCMC METHODS FOR DIFFUSION BRIDGES / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Hybrid Monte Carlo on Hilbert spaces / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: On the small balls problem for equivalent Gaussian measures / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Q3415147 / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Markov chains for exploring posterior distributions. (With discussion) / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Q5614192 / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: MCMC methods for functions: modifying old algorithms to make them faster / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Pointwise ergodic theorems with rate and application to the CLT for Markov chains / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Besov priors for Bayesian inverse problems / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Uncertainty Quantification and Weak Approximation of an Elliptic Inverse Problem / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Stochastic Equations in Infinite Dimensions / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Geometric bounds for eigenvalues of Markov chains / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Error bounds for Metropolis-Hastings algorithms applied to perturbations of Gaussian measures in high dimensions / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Q4272795 / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Annealing Markov Chain Monte Carlo with Applications to Ancestral Inference / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: A simple framework to justify linear response theory / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Asymptotic coupling and a general form of Harris' theorem with applications to stochastic delay equations / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Analysis of SPDEs arising in path sampling. II: The nonlinear case / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Monte Carlo sampling methods using Markov chains and their applications / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Bayesian Nonparametrics / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Curvature, concentration and error estimates for Markov chain Monte Carlo / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Central limit theorem for additive functionals of reversible Markov processes and applications to simple exclusions / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Central limit theorem for Markov processes with spectral gap in the Wasserstein metric / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Discretization-invariant Bayesian inversion and Besov space priors / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Rigorous confidence bounds for MCMC under a geometric drift condition / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: CLTs and asymptotic variance of time-sampled Markov chains / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Bounds on the L 2 Spectrum for Markov Chains and Markov Processes: A Generalization of Cheeger's Inequality / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Q4827935 / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Monte Carlo strategies in scientific computing. / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Random walks in a convex body and an improved volume algorithm / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Diffusion limits of the random walk Metropolis algorithm in high dimensions / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Markov chains and stochastic stability / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: A splitting technique for Harris recurrent Markov chains / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Q4828566 / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Geometric convergence and central limit theorems for multidimensional Hastings and Metropolis algorithms / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Weak Poincaré inequalities and \(L^2\)-convergence rates of Markov semigroups / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Explicit error bounds for Markov chain Monte Carlo / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Sparse deterministic approximation of Bayesian inverse problems / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Approximate counting, uniform generation and rapidly mixing Markov chains / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Inverse problems: A Bayesian perspective / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: A note on Metropolis-Hastings kernels for general state spaces / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Functional inequalities for the decay of sub-Markov semigroups / rank | |||
Normal rank | |||
links / mardi / name | links / mardi / name | ||
Latest revision as of 07:38, 9 July 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Spectral gaps for a Metropolis-Hastings algorithm in infinite dimensions |
scientific article |
Statements
Spectral gaps for a Metropolis-Hastings algorithm in infinite dimensions (English)
0 references
21 November 2014
0 references
It is well known that the most widely used method for general target measures are Markov chain Monte Carlo (MCMC) algorithms. The main aim of this article is to study the complexity of certain sampling algorithms in high dimensions, more precisely, the Metropolis-Hastings algorithm in infinite dimensions. The computational cost of such an algorithm is the number of necessary steps \(\times\) cost of a step. While for most algorithms the cost of one step grows with the dimension, a major result of this paper is an lgorithm which, when applied to measures, defines via a finite-dimensional approximation of a measure defined by a density with respect to a Gaussian random field, requires a number of steps independent of the dimension in order to achieve a given level of accuracy. The authors consider Metropolis-Hastings MCMC methods and focus on cases when the proposal is either a Gaussian random walk (RWM) for which the covariance equals to that of a reference measure or an Orstain-Uhlenbeck proposal (pCN) for which the reference measure is invariant. Among the main results belong: {\parindent=6mm \begin{itemize}\item[{\(\bullet\)}] Description of \(RWM\) and \(pCN\) algorithms and study of their properties, both positive and negative ones. \item[{\(\bullet\)}] Statement of the weak Harris theorem and a discussion of the relationship between exponential convergence in a Wasserstein distance and \(L^{2}_{\mu}\)-spectral gaps. \item[{\(\bullet\)}] Formal confirmation of a conjecture that \(pCN\) has a convergence rate that is independent of the dimension while \(RWM\) has undesirable dimension dependent properties. \end{itemize}} The results are presented for separable Hilbert spaces, but in fact they do also hold on an arbitrary Banach space.
0 references
Wasserstein spectral gaps
0 references
\(L^{2}\)-spectral gaps
0 references
Markov chain Monte Carlo in infinite dimensions
0 references
weak Harris theorem
0 references
random walk
0 references
Gaussian random field
0 references
sampling algorithm
0 references
Metropolis-Hastings algorithm
0 references
0 references
0 references
0 references
0 references
0 references
0 references