Discontinuous Hamiltonian Monte Carlo for discrete parameters and discontinuous likelihoods

DOI10.1093/BIOMET/ASZ083MaRDI QIDQ5113016zbMATH OpenOpenAlexWikidataFDO

Authors Akihiko Nishimura, David Dunson, Jianfeng Lu

Publication date 9 June 2020

Published in Biometrika (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1705.08510

zbMATH Keywords

Bayesian inferencegeometric numerical integrationMarkov chain Monte Carlomeasure-valued differential equation

Mathematics Subject Classification ID

Computational methods for problems pertaining to statistics (62-08) Bayesian inference (62F15) Monte Carlo methods (65C05)

Abstract: Hamiltonian Monte Carlo has emerged as a standard tool for posterior computation. In this article, we present an extension that can efficiently explore target distributions with discontinuous densities. Our extension in particular enables efficient sampling from ordinal parameters though embedding of probability mass functions into continuous spaces. We motivate our approach through a theory of discontinuous Hamiltonian dynamics and develop a corresponding numerical solver. The proposed solver is the first of its kind, with a remarkable ability to exactly preserve the Hamiltonian. We apply our algorithm to challenging posterior inference problems to demonstrate its wide applicability and competitive performance.

Recommendations

• The no-U-turn sampler: adaptively setting path lengths in Hamiltonian Monte Carlo
• Hamiltonian Monte Carlo with energy conserving subsampling
• Hamiltonian Monte Carlo sampling in Bayesian empirical likelihood computation
• MCMC using Hamiltonian dynamics
• Riemann manifold Langevin and Hamiltonian Monte Carlo methods. With discussion and authors' reply

Cited in

(21)

• Densities of almost surely terminating probabilistic programs are differentiable almost everywhere
• Modified Hamiltonian Monte Carlo for Bayesian inference
• Efficient sequential Monte Carlo algorithms for integrated population models
• Direct sampling with a step function
• Continualization of probabilistic programs with correction
• Bayesian inference of scaled versus fractional Brownian motion
• Zig-Zag Sampling for Discrete Structures and Nonreversible Phylogenetic MCMC
• The Block-Poisson Estimator for Optimally Tuned Exact Subsampling MCMC
• Symmetrically processed splitting integrators for enhanced Hamiltonian Monte Carlo sampling
• Hamiltonian Monte Carlo with energy conserving subsampling
• Transport Monte Carlo: High-Accuracy Posterior Approximation via Random Transport
• A New Optimality Property of Strang’s Splitting
• Sampling algorithms in statistical physics: a guide for statistics and machine learning
• Super-efficient exact Hamiltonian Monte Carlo for the von Mises distribution
• Numerical generalized randomized HMC processes for restricted domains
• Zigzag Path Connects Two Monte Carlo Samplers: Hamiltonian Counterpart to a Piecewise Deterministic Markov Process
• PDMP Monte Carlo methods for piecewise smooth densities
• A generalized discontinuous Hamilton Monte Carlo for transdimensional sampling
• Parallel computations for Metropolis Markov chains with Picard maps
• Graph-accelerated Markov chain Monte Carlo using approximate samples
• XDNUTS

This page was built for publication: Discontinuous Hamiltonian Monte Carlo for discrete parameters and discontinuous likelihoods

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5113016)