Geometric integrators and the Hamiltonian Monte Carlo method

DOI10.1017/S0962492917000101zbMATH Open1431.65004arXiv1711.05337OpenAlexW2768877132WikidataQ129871765 ScholiaQ129871765MaRDI QIDQ5230516FDOQ5230516

Jesús María Sanz-Serna, Nawaf Bou-Rabee

Publication date: 28 August 2019

Published in: Acta Numerica (Search for Journal in Brave)

Abstract: This paper surveys in detail the relations between numerical integration and the Hamiltonian (or hybrid) Monte Carlo method (HMC). Since the computational cost of HMC mainly lies in the numerical integrations, these should be performed as efficiently as possible. However, HMC requires methods that have the geometric properties of being volume-preserving and reversible, and this limits the number of integrators that may be used. On the other hand, these geometric properties have important quantitative implications on the integration error, which in turn have an impact on the acceptance rate of the proposal. While at present the velocity Verlet algorithm is the method of choice for good reasons, we argue that Verlet can be improved upon. We also discuss in detail the behavior of HMC as the dimensionality of the target distribution increases.

Full work available at URL: https://arxiv.org/abs/1711.05337

Mathematics Subject Classification ID

Monte Carlo methods (65C05) Research exposition (monographs, survey articles) pertaining to numerical analysis (65-02) Numerical integration (65D30)

Cites Work

• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Solving Ordinary Differential Equations II
• GSHMC: An efficient method for molecular simulation
• The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo
• Slice sampling. (With discussions and rejoinder)
• Monte Carlo strategies in scientific computing.
• Exponential convergence of Langevin distributions and their discrete approximations
• Weak convergence and optimal scaling of random walk Metropolis algorithms
• Markov chains for exploring posterior distributions. (With discussion)
• An improved acceptance procedure for the hybrid Monte Carlo algorithm
• Analysis of a nonreversible Markov chain sampler.
• Delayed rejection in reversible jump Metropolis-Hastings.
• Sampling-Based Approaches to Calculating Marginal Densities
• Stochastic Relaxation, Gibbs Distributions, and the Bayesian Restoration of Images
• Monte Carlo sampling methods using Markov chains and their applications
• Stochastic simulation: Algorithms and analysis
• Geometric convergence and central limit theorems for multidimensional Hastings and Metropolis algorithms
• Optimal Scaling of Discrete Approximations to Langevin Diffusions
• Equi-energy sampler with applications in statistical inference and statistical mechanics
• Equation of State Calculations by Fast Computing Machines
• Riemann Manifold Langevin and Hamiltonian Monte Carlo Methods
• Annealing Markov Chain Monte Carlo with Applications to Ancestral Inference
• The BUGS book. A practical introduction to Bayesian analysis
• Symplectic Runge-Kutta schemes for adjoint equations, automatic differentiation, optimal control, and more
• Inverse problems: a Bayesian perspective
• Solving Ordinary Differential Equations I
• The Markov chain Monte Carlo revolution
• Symplectic Geometric Algorithms for Hamiltonian Systems
• Stochastic Equations in Infinite Dimensions
• Theoretical and numerical comparison of some sampling methods for molecular dynamics
• Time-reversal symmetry in dynamical systems: a survey
• Word series for dynamical systems and their numerical integrators
• Simulating Hamiltonian Dynamics
• On the Numerical Integration of Ordinary Differential Equations by Symmetric Composition Methods
• Real-Parameter Evolutionary Monte Carlo With Applications to Bayesian Mixture Models
• On the Scope of the Method of Modified Equations
• Reflection couplings and contraction rates for diffusions
• An introduction to MCMC for machine learning
• The ensemble Kalman filter for combined state and parameter estimation
• On Metropolis-Hastings algorithms with delayed rejection
• Optimal tuning of the hybrid Monte Carlo algorithm
• Practical construction of modified Hamiltonians
• MCMC METHODS FOR DIFFUSION BRIDGES
• Hybrid Monte Carlo on Hilbert spaces
• A patch that imparts unconditional stability to explicit integrators for Langevin-like equations
• Nonasymptotic mixing of the MALA algorithm
• Shadow hybrid Monte Carlo: an efficient propagator in phase space of macromolecules
• Cost of the generalised hybrid Monte Carlo algorithm for free field theory
• Molecular modeling and simulation. An interdisciplinary guide
• Geometric Numerical Integration
• Pathwise accuracy and ergodicity of metropolized integrators for SDEs
• Error Growth in the Numerical Integration of Periodic Orbits, with Application to Hamiltonian and Reversible Systems
• Numerical Methods for Ordinary Differential Equations
• Computational discovery of gene regulatory binding motifs: a Bayesian perspective
• Invariant measures of stochastic partial differential equations and conditioned diffusions
• The Development of Variable-Step Symplectic Integrators, with Application to the Two-Body Problem
• Accurate partial difference methods. I: Linear Cauchy problems
• Metropolis Integration Schemes for Self-Adjoint Diffusions
• Error analysis of the transport properties of Metropolized schemes
• Instabilities and Inaccuracies in the Integration of Highly Oscillatory Problems
• A concise introduction to geometric numerical integration
• Order conditions for numerical integrators obtained by composing simpler integrators
• Randomized Hamiltonian Monte Carlo
• Error growth in the numerical integration of periodic orbits by multistep methods, with application to reversible systems
• Numerical Integrators for the Hybrid Monte Carlo Method
• Introduction to Uncertainty Quantification
• Markov Chain Monte Carlo and Numerical Differential Equations
• Extra chance generalized hybrid Monte Carlo
• Couplings and quantitative contraction rates for Langevin dynamics
• Numerical Methods for Ordinary Differential Equations
• Computational Biology: Toward Deciphering Gene Regulatory Information in Mammalian Genomes
• Adaptive multi-stage integrators for optimal energy conservation in molecular simulations
• Palindromic 3-stage splitting integrators, a roadmap
• A technique for studying strong and weak local errors of splitting stochastic integrators
• On the application of improved symplectic integrators in Hamiltonian Monte Carlo
• Coupling and convergence for Hamiltonian Monte Carlo

Cited In (48)

• An Hermite-Obreshkov method for 2nd-order linear initial-value problems for ODE. With special attention paid to the Mathieu equation
• On explicit \(L^2\)-convergence rate estimate for piecewise deterministic Markov processes in MCMC algorithms
• On the accept-reject mechanism for Metropolis-Hastings algorithms
• Mixing time guarantees for unadjusted Hamiltonian Monte Carlo
• GPU-accelerated particle methods for evaluation of sparse observations for inverse problems constrained by diffusion PDEs
• On the geometric ergodicity of Hamiltonian Monte Carlo
• Irreducibility and geometric ergodicity of Hamiltonian Monte Carlo
• Title not available (Why is that?)
• Improving sampling accuracy of stochastic gradient MCMC methods via non-uniform subsampling of gradients
• A New Optimality Property of Strang’s Splitting
• Continuum limit and preconditioned Langevin sampling of the path integral molecular dynamics
• Geometrical Monte Carlo method and its modifications
• Adaptive parameters tuning based on energy-preserving splitting integration for Hamiltonian Monte Carlo method
• Two-scale coupling for preconditioned Hamiltonian Monte Carlo in infinite dimensions
• Multi-stage splitting integrators for sampling with modified Hamiltonian Monte Carlo methods
• Computing quantum dynamics in the semiclassical regime
• Mixing of Metropolis-adjusted Markov chains via couplings: the high acceptance regime
• Splitting methods for differential equations
• Choice of integrator in the hybrid Monte Carlo algorithm
• Coupling and convergence for Hamiltonian Monte Carlo
• Truncated log-concave sampling for convex bodies with reflective Hamiltonian Monte Carlo
• Learning variational autoencoders via MCMC speed measures
• Log-density gradient covariance and automatic metric tensors for Riemann manifold Monte Carlo methods
• Adaptive tuning of Hamiltonian Monte Carlo within sequential Monte Carlo
• On the application of improved symplectic integrators in Hamiltonian Monte Carlo
• Geometric Integration of Measure-Preserving Flows for Sampling
• Symmetrically Processed Splitting Integrators for Enhanced Hamiltonian Monte Carlo Sampling
• B-methods for the numerical solution of evolution problems with blow-up solutions. II: Splitting methods
• Uncertainty quantification for random Hamiltonian systems by using polynomial expansions and geometric integrators
• Convergence of unadjusted Hamiltonian Monte Carlo for mean-field models
• Mixing rates for Hamiltonian Monte Carlo algorithms in finite and infinite dimensions
• A Bayesian Approach to Estimating Background Flows from a Passive Scalar
• Tuning diagonal scale matrices for HMC
• Connecting the Dots: Numerical Randomized Hamiltonian Monte Carlo with State-Dependent Event Rates
• Motion equations in a Kerr–Newman–de Sitter spacetime: some methods of integration and application to black holes shadowing in Scilab
• Geometric adaptive Monte Carlo in random environment
• Applying kriging proxies for Markov chain Monte Carlo in reservoir simulation
• Geometric numerical integration. Abstracts from the workshop held March 28 -- April 3, 2021 (hybrid meeting)
• Adaptive dimension reduction to accelerate infinite-dimensional geometric Markov chain Monte Carlo
• Geodesic Lagrangian Monte Carlo over the space of positive definite matrices: with application to Bayesian spectral density estimation
• Tuning Symplectic Integrators is Easy and Worthwhile
• Cores for piecewise-deterministic Markov processes used in Markov chain Monte Carlo
• Data assimilation: The Schrödinger perspective
• Sampling algorithms in statistical physics: a guide for statistics and machine learning
• Asymptotic bias of inexact Markov chain Monte Carlo methods in high dimension
• Split Hamiltonian Monte Carlo revisited
• HMC: reducing the number of rejections by not using leapfrog and some results on the acceptance rate
• Adaptive multi-stage integration schemes for Hamiltonian Monte Carlo

This page was built for publication: Geometric integrators and the Hamiltonian Monte Carlo method

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5230516)