Scalable Control Variates for Monte Carlo Methods Via Stochastic Optimization

DOI10.1007/978-3-030-98319-2_10arXiv2006.07487OpenAlexW3034678031MaRDI QIDQ6154293FDOQ6154293

Authors: Shijing Si, Chris J. Oates, Andrew B. Duncan, Lawrence Carin, François-Xavier Briol 

Publication date: 14 February 2024

Published in: Springer Proceedings in Mathematics & Statistics (Search for Journal in Brave)

Abstract: Control variates are a well-established tool to reduce the variance of Monte Carlo estimators. However, for large-scale problems including high-dimensional and large-sample settings, their advantages can be outweighed by a substantial computational cost. This paper considers control variates based on Stein operators, presenting a framework that encompasses and generalizes existing approaches that use polynomials, kernels and neural networks. A learning strategy based on minimising a variational objective through stochastic optimization is proposed, leading to scalable and effective control variates. Novel theoretical results are presented to provide insight into the variance reduction that can be achieved, and an empirical assessment, including applications to Bayesian inference, is provided in support.

Full work available at URL: https://arxiv.org/abs/2006.07487

Recommendations

• Control Functionals for Monte Carlo Integration
• Diffusion approximations and control variates for MCMC
• Variance reduction for Markov chains with application to MCMC
• Control Variates for Estimation Based on Reversible Markov Chain Monte Carlo Samplers
• Control variates for stochastic gradient MCMC

zbMATH Keywords

control variatesvariance reductionMonte Carlo methodsStein's method

Mathematics Subject Classification ID

Monte Carlo methods (65C05) Numerical mathematical programming methods (65K05) Stochastic programming (90C15)

Cites Work

• Bayesian calibration of computer models. (With discussion)
• Probabilistic integration: a role in statistical computation?
• Approximating Martingales for Variance Reduction in Markov Process Simulation
• Zero variance Markov chain Monte Carlo for Bayesian estimators
• Bayes-Hermite quadrature
• Fundamentals of Stein's method
• Title not available (Why is that?)
• Control variates for quasi-Monte Carlo (with comments and rejoinder)
• Parametric Stein operators and variance bounds
• Monte Carlo integration with a growing number of control variates
• Control Functionals for Monte Carlo Integration
• Variance Reduction for Simulated Diffusions
• On the Poisson equation and diffusion approximation. I
• Control Variates for Estimation Based on Reversible Markov Chain Monte Carlo Samplers
• Variance reduction techniques for gradient estimates in reinforcement learning
• Title not available (Why is that?)
• Variance reduction through smoothing and control variates for Markov chain simulations
• Exploiting multi-core architectures for reduced-variance estimation with intractable likelihoods
• Control Variates for the Metropolis–Hastings Algorithm
• Zero variance differential geometric Markov chain Monte Carlo algorithms
• Measuring sample quality with diffusions
• Variance reduction for Markov chains with application to MCMC
• Convergence rates for a class of estimators based on Stein's method
• Variance reduction in Monte Carlo estimators via empirical variance minimization
• Control variates for stochastic gradient MCMC

Cited In (2)

• Diffusion approximations and control variates for MCMC
• Regularized zero-variance control variates

This page was built for publication: Scalable Control Variates for Monte Carlo Methods Via Stochastic Optimization

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6154293)