Hamiltonian Monte Carlo Without Detailed Balance
arXiv:1409.5191
Abstract
We present a method for performing Hamiltonian Monte Carlo that largely eliminates sample rejection for typical hyperparameters. In situations that would normally lead to rejection, instead a longer trajectory is computed until a new state is reached that can be accepted. This is achieved using Markov chain transitions that satisfy the fixed point equation, but do not satisfy detailed balance. The resulting algorithm significantly suppresses the random walk behavior and wasted function evaluations that are typically the consequence of update rejection. We demonstrate a greater than factor of two improvement in mixing time on three test problems. We release the source code as Python and MATLAB packages.
Accepted conference submission to ICML 2014 and also featured in a special edition of JMLR. Since updated to include additional literature citations
References in corpus (5)
- Markov Chain Monte Carlo Method without Detailed Balance
- Extra Chance Generalized Hybrid Monte Carlo
- Improving the Asymptotic Performance of Markov Chain Monte-Carlo by Inserting Vortices
- Hamiltonian Annealed Importance Sampling for partition function estimation
- Hamiltonian Monte Carlo with Reduced Momentum Flips
Cited by in corpus (17)
- On the Geometric Ergodicity of Hamiltonian Monte Carlo
- Compressible Generalized Hybrid Monte Carlo
- Modified Hamiltonian Monte Carlo for Bayesian inference
- The Geometric Foundations of Hamiltonian Monte Carlo
- Identifying the Optimal Integration Time in Hamiltonian Monte Carlo
- Generalizing Hamiltonian Monte Carlo with Neural Networks
- A general perspective on the Metropolis-Hastings kernel
- Peskun-Tierney ordering for Markov chain and process Monte Carlo: beyond the reversible scenario
- A Neural Network MCMC sampler that maximizes Proposal Entropy
- Delayed rejection Hamiltonian Monte Carlo for sampling multiscale distributions
- Nonreversible MCMC from conditional invertible transforms: a complete recipe with convergence guarantees
- Bayesian Neural Networks at Finite Temperature
- A Common Derivation for Markov Chain Monte Carlo Algorithms with Tractable and Intractable Targets
- A Unifying and Canonical Description of Measure-Preserving Diffusions
- A Markov Jump Process for More Efficient Hamiltonian Monte Carlo
- Involutive MCMC: a Unifying Framework
- Hamiltonian Dynamics with Non-Newtonian Momentum for Rapid Sampling