Optimal tuning of the hybrid Monte Carlo algorithm

DOI10.3150/12-BEJ414MaRDI QIDQ2435211zbMATH OpenOpenAlexFDO

Authors Alexandros Beskos, Natesh S. Pillai, Gareth O. Roberts, A. M. Stuart, Jesús María Sanz-Serna

Publication date 4 February 2014

Published in Bernoulli (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1001.4460

zbMATH Keywords

Hamiltonian dynamicshigh dimensionsleapfrog schemeoptimal acceptance probabilitysquared jumping distance

Mathematics Subject Classification ID

Computational methods in Markov chains (60J22)

Abstract: We investigate the properties of the Hybrid Monte-Carlo algorithm (HMC) in high dimensions. HMC develops a Markov chain reversible w.r.t. a given target distribution $Pi$ by using separable Hamiltonian dynamics with potential $-logPi$. The additional momentum variables are chosen at random from the Boltzmann distribution and the continuous-time Hamiltonian dynamics are then discretised using the leapfrog scheme. The induced bias is removed via a Metropolis-Hastings accept/reject rule. In the simplified scenario of independent, identically distributed components, we prove that, to obtain an $mathcalO(1)$ acceptance probability as the dimension $d$ of the state space tends to $infty$, the leapfrog step-size $h$ should be scaled as $h=limes{d}^{-1/4}$. Therefore, in high dimensions, HMC requires $mathcalO(d^{1/4})$ steps to traverse the state space. We also identify analytically the asymptotically optimal acceptance probability, which turns out to be 0.651 (to three decimal places). This is the choice which optimally balances the cost of generating a proposal, which {em decreases} as $l$ increases, against the cost related to the average number of proposals required to obtain acceptance, which {em increases} as $l$ increases.

Recommendations

• scientific article; zbMATH DE number 1163492
• Accelerating the hybrid Monte Carlo algorithm
• An Optimal Algorithm for Monte Carlo Estimation
• Adaptive optimization of the Monte-Carlo method
• Monte Carlo optimization
• Choice of integrator in the hybrid Monte Carlo algorithm
• Optimization of randomized Monte Carlo algorithms for solving problems with random parameters
• Optimal Monte Carlo integration with fixed relative precision
• Techniques for Monte Carlo Optimizing

Cites work

• A comparison of generalized hybrid Monte Carlo methods with and without momentum flip
• Accelerated Monte Carlo for optimal estimation of time series
• Adaptively scaling the Metropolis algorithm using expected squared jumped distance
• An Introduction to Stein's Method
• Analysis of a nonreversible Markov chain sampler.
• Bayesian learning for neural networks
• Diffusion limits of the random walk Metropolis algorithm in high dimensions
• Exponential convergence of Langevin distributions and their discrete approximations
• Geometric Numerical Integration
• scientific article; zbMATH DE number 4041188 (Why is no real title available?)
• scientific article; zbMATH DE number 3723610 (Why is no real title available?)
• scientific article; zbMATH DE number 702482 (Why is no real title available?)
• scientific article; zbMATH DE number 1405381 (Why is no real title available?)
• Hybrid Monte Carlo on Hilbert spaces
• Manifold stochastic dynamics for Bayesian learning
• MCMC METHODS FOR DIFFUSION BRIDGES
• MCMC methods for sampling function space
• Monte Carlo strategies in scientific computing.
• Optimal acceptance rates for Metropolis algorithms: Moving beyond 0.234
• Optimal scaling and diffusion limits for the Langevin algorithm in high dimensions
• Optimal scaling for various Metropolis-Hastings algorithms.
• Optimal Scaling of Discrete Approximations to Langevin Diffusions
• Optimal scalings for local Metropolis-Hastings chains on nonproduct targets in high dimensions
• Scaling Limits for the Transient Phase of Local Metropolis–Hastings Algorithms
• Shadow hybrid Monte Carlo: an efficient propagator in phase space of macromolecules
• Simulating Hamiltonian Dynamics
• Speeding up the hybrid Monte Carlo algorithm for dynamical fermions
• Theoretical and numerical comparison of some sampling methods for molecular dynamics
• Weak convergence and optimal scaling of random walk Metropolis algorithms
• Weak convergence of Metropolis algorithms for non-I.I.D. target distributions

Cited in

(only showing first 100 items - show all)

• An improved acceptance procedure for the hybrid Monte Carlo algorithm
• Merging MCMC subposteriors through Gaussian-process approximations
• Designing simple and efficient Markov chain Monte Carlo proposal kernels
• Hierarchical models: local proposal variances for RWM-within-Gibbs and MALA-within-Gibbs
• Adaptive multi-stage integrators for optimal energy conservation in molecular simulations
• Randomized Hamiltonian Monte Carlo
• Palindromic 3-stage splitting integrators, a roadmap
• Constrained Hamiltonian Monte Carlo in BEKK GARCH with targeting
• Modified Cholesky Riemann manifold Hamiltonian Monte Carlo: exploiting sparsity for fast sampling of high-dimensional targets
• Simulating Coulomb and log-gases with hybrid Monte Carlo algorithms
• Leave Pima Indians alone: binary regression as a benchmark for Bayesian computation
• Optimal scaling and diffusion limits for the Langevin algorithm in high dimensions
• Weak convergence and optimal tuning of the reversible jump algorithm
• Two-scale coupling for preconditioned Hamiltonian Monte Carlo in infinite dimensions
• Recycling intermediate steps to improve Hamiltonian Monte Carlo
• Randomized Hamiltonian Monte Carlo as scaling limit of the bouncy particle sampler and dimension-free convergence rates
• Spatial voting models in circular spaces: a case study of the U.S. House of Representatives
• Adaptive random neighbourhood informed Markov chain Monte Carlo for high-dimensional Bayesian variable selection
• Semi-supervised nonparametric Bayesian modelling of spatial proteomics
• Split Hamiltonian Monte Carlo revisited
• Mixing rates for Hamiltonian Monte Carlo algorithms in finite and infinite dimensions
• HMC: reducing the number of rejections by not using leapfrog and some results on the acceptance rate
• Optimal scaling of random walk Metropolis algorithms using Bayesian large-sample asymptotics
• An adaptive multiple-try Metropolis algorithm
• Copula multivariate GARCH model with constrained Hamiltonian Monte Carlo
• Optimal scaling of random-walk Metropolis algorithms on general target distributions
• Markov chain Monte Carlo algorithms with sequential proposals
• Localization for MCMC: sampling high-dimensional posterior distributions with local structure
• Leveraging Bayesian analysis to improve accuracy of approximate models
• Mixing of Hamiltonian Monte Carlo on strongly log-concave distributions: continuous dynamics
• Parameter estimation in stochastic differential equations with Markov chain Monte Carlo and non-linear Kalman filtering
• Hierarchical models and tuning of random walk Metropolis algorithms
• Random walk Metropolis algorithm in high dimension with non-Gaussian target distributions
• Modified Hamiltonian Monte Carlo for Bayesian inference
• On the geometric ergodicity of Hamiltonian Monte Carlo
• Optimal scaling for the transient phase of the random walk Metropolis algorithm: the mean-field limit
• Rejoinder on: ``Some recent work on multivariate Gaussian Markov random fields
• On the stability of sequential Monte Carlo methods in high dimensions
• Adaptive thermostats for noisy gradient systems
• Adaptive step size selection for Hessian-based manifold Langevin samplers
• Accelerated dimension-independent adaptive metropolis
• Thread-parallel anisotropic mesh adaptation
• MCMC using Hamiltonian dynamics
• MALA-within-Gibbs samplers for high-dimensional distributions with sparse conditional structure
• FEM-based discretization-invariant MCMC methods for PDE-constrained Bayesian inverse problems
• Noisy Hamiltonian Monte Carlo for Doubly Intractable Distributions
• Multiple-time-stepping generalized hybrid Monte Carlo methods
• Optimum Monte Carlo simulations: some exact results
• scientific article; zbMATH DE number 1560250 (Why is no real title available?)
• On the application of improved symplectic integrators in Hamiltonian Monte Carlo
• Optimal scaling for the transient phase of Metropolis Hastings algorithms: the longtime behavior
• Fast mixing of Metropolized Hamiltonian Monte Carlo: benefits of multi-step gradients
• Kinetic energy choice in Hamiltonian/hybrid Monte Carlo
• Forward Event-Chain Monte Carlo: Fast Sampling by Randomness Control in Irreversible Markov Chains
• Adaptive multiple importance sampling for Gaussian processes
• Bayesian computation for Log-Gaussian Cox processes: a comparative analysis of methods
• Symmetrically processed splitting integrators for enhanced Hamiltonian Monte Carlo sampling
• Maximum Conditional Entropy Hamiltonian Monte Carlo Sampler
• Asymptotic variance for random walk Metropolis chains in high dimensions: logarithmic growth via the Poisson equation
• Hamiltonian Monte Carlo with energy conserving subsampling
• Geometric integrators and the Hamiltonian Monte Carlo method
• Finite element model updating using Hamiltonian Monte Carlo techniques
• Rejoinder: Geodesic Monte Carlo on embedded manifolds
• Skew Brownian motion and complexity of the ALPS algorithm
• Cost of the generalised hybrid Monte Carlo algorithm for free field theory
• Bayesian computation: a summary of the current state, and samples backwards and forwards
• MCMC methods for functions: modifying old algorithms to make them faster
• Tuning the hybrid Monte Carlo algorithm using molecular dynamics forces' variances
• Bayesian elastic net based on empirical likelihood
• Non-reversible guided Metropolis kernel
• Adaptive multi-stage integration schemes for Hamiltonian Monte Carlo
• System identification using autoregressive Bayesian neural networks with nonparametric noise models
• Convergence of unadjusted Hamiltonian Monte Carlo for mean-field models
• Concave-Convex PDMP-based Sampling
• The Apogee to Apogee Path Sampler
• Adaptive tuning of Hamiltonian Monte Carlo within sequential Monte Carlo
• Dimension-free mixing times of Gibbs samplers for Bayesian hierarchical models
• Asymptotic bias of inexact Markov chain Monte Carlo methods in high dimension
• An embedded diachronic sense change model with a case study from ancient Greek
• Randomized physics-informed machine learning for uncertainty quantification in high-dimensional inverse problems
• Tuning diagonal scale matrices for HMC
• Posterior computation with the Gibbs zig-zag sampler
• Sequential Kalman tuning of the t-preconditioned Crank-Nicolson algorithm: efficient, adaptive and gradient-free inference for Bayesian inverse problems
• Nested \(\widehat{R}\): assessing the convergence of Markov chain Monte Carlo when running many short chains
• Optimal design of the Barker proposal and other locally balanced Metropolis-Hastings algorithms
• Sparse Bayesian multidimensional scaling(s)
• Scalable couplings for the random walk Metropolis algorithm
• Sampling parameters of ordinary differential equations with constrained Langevin dynamics
• Accelerate Langevin sampling with birth-death process and exploration component
• A splitting Hamiltonian Monte Carlo method for efficient sampling
• Scaling of piecewise deterministic Monte Carlo for anisotropic targets
• Multi-fidelity No-U-Turn Sampling
• Delayed rejection Hamiltonian Monte Carlo for sampling multiscale distributions
• Time series analysis of fMRI data: spatial modelling and Bayesian computation
• A flexible choice of critical constants for the improved hybrid Hochberg-Hommel procedure
• Provably Efficient Posterior Sampling for Sparse Linear Regression via Measure Decomposition
• A generalized discontinuous Hamilton Monte Carlo for transdimensional sampling
• Exact simulation of diffusions and improved algorithms for log-concave sampling
• Extra chance generalized hybrid Monte Carlo
• Title not available (Why is no real title available?)

This page was built for publication: Optimal tuning of the hybrid Monte Carlo algorithm

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2435211)