Variance reduction for Metropolis-Hastings samplers
From MaRDI portal
Publication:2104009
Abstract: We introduce a general framework that constructs estimators with reduced variance for random walk Metropolis and Metropolis-adjusted Langevin algorithms. The resulting estimators require negligible computational cost and are derived in a post-process manner utilising all proposal values of the Metropolis algorithms. Variance reduction is achieved by producing control variates through the approximate solution of the Poisson equation associated with the target density of the Markov chain. The proposed method is based on approximating the target density with a Gaussian and then utilising accurate solutions of the Poisson equation for the Gaussian case. This leads to an estimator that uses two key elements: (i) a control variate from the Poisson equation that contains an intractable expectation under the proposal distribution, (ii) a second control variate to reduce the variance of a Monte Carlo estimate of this latter intractable expectation. Simulated data examples are used to illustrate the impressive variance reduction achieved in the Gaussian target case and the corresponding effect when target Gaussianity assumption is violated. Real data examples on Bayesian logistic regression and stochastic volatility models verify that considerable variance reduction is achieved with negligible extra computational cost.
Recommendations
- Variance reduction of estimators arising from Metropolis-Hastings algorithms
- Variance reduction using nonreversible Langevin samplers
- Optimal variance reduction for Markov chain Monte Carlo
- Variance reduction for Markov chains with application to MCMC
- Variance reduction in Monte Carlo estimators via empirical variance minimization
- Variance reduction for Fokker-Planck based particle Monte Carlo schemes
- A modified conditional Metropolis-Hastings sampler
- Note on the Sampling Distribution for the Metropolis-Hastings Algorithm
- Variance reduction for irreversible Langevin samplers and diffusion on graphs
Cites work
- A New Multinomial Model and a Zero Variance Estimation
- A Riemann-Stein kernel method
- All of Nonparametric Statistics
- Ancillarity-sufficiency interweaving strategy (ASIS) for boosting MCMC estimation of stochastic volatility models
- Asymptotic variance for random walk Metropolis chains in high dimensions: logarithmic growth via the Poisson equation
- Batch means and spectral variance estimators in Markov chain Monte Carlo
- Control functionals for Monte Carlo integration
- Control Techniques for Complex Networks
- Control Variates for Estimation Based on Reversible Markov Chain Monte Carlo Samplers
- Control Variates for the Metropolis–Hastings Algorithm
- Convergence rates for a class of estimators based on Stein's method
- Geometric ergodicity and hybrid Markov chains
- scientific article; zbMATH DE number 1999206 (Why is no real title available?)
- scientific article; zbMATH DE number 1560252 (Why is no real title available?)
- scientific article; zbMATH DE number 2148859 (Why is no real title available?)
- Improving Stochastic Relaxation for Gussian Random Fields
- Methods of reducing sample size in Monte Carlo computations
- Multiprocess parallel antithetic coupling for backward and forward Markov chain Monte Carlo
- NOTES ON BIAS IN ESTIMATION
- On the Poisson equation for Metropolis-Hastings chains
- Rank-normalization, folding, and localization: an improved \(\widehat{R}\) for assessing convergence of MCMC (with Discussion)
- Riemann manifold Langevin and Hamiltonian Monte Carlo methods. With discussion and authors' reply
- Sampling-Based Approaches to Calculating Marginal Densities
- Semi-exact control functionals from Sard’s method
- Stochastic models, statistical methods, and algorithms in image analysis. Proceedings of the special year on image analysis, held in Rome, Italy, 1990
- Stochastic Volatility: Likelihood Inference and Comparison with ARCH Models
- The Convergence of a Class of Double-rank Minimization Algorithms 1. General Considerations
- Variance reduction for Markov chains with application to MCMC
- Variance reduction through smoothing and control variates for Markov chain simulations
- Zero variance differential geometric Markov chain Monte Carlo algorithms
- Zero variance Markov chain Monte Carlo for Bayesian estimators
Cited in
(17)- On the Poisson equation for Metropolis-Hastings chains
- Subsampling the Gibbs sampler: variance reduction
- Variance reduction for additive functionals of Markov chains via martingale representations
- Variance bounding of delayed-acceptance kernels
- Variance reduction for Markov chains with application to MCMC
- Variance reduction using nonreversible Langevin samplers
- Control Variates for Estimation Based on Reversible Markov Chain Monte Carlo Samplers
- Robust Variance Reduction for Random Walk Methods
- Variance reduction for dependent sequences with applications to stochastic gradient MCMC
- Asymptotic variance for random walk Metropolis chains in high dimensions: logarithmic growth via the Poisson equation
- Control Variates for the Metropolis–Hastings Algorithm
- Diffusion approximations and control variates for MCMC
- Solving the Poisson equation using coupled Markov chains
- Unbiased and consistent nested sampling via sequential Monte Carlo
- Variance reduction of estimators arising from Metropolis-Hastings algorithms
- Zero variance differential geometric Markov chain Monte Carlo algorithms
- Zero variance Markov chain Monte Carlo for Bayesian estimators
This page was built for publication: Variance reduction for Metropolis-Hastings samplers
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2104009)
