Irreversible Langevin MCMC on Lie Groups

03/21/2019
by   Alexis Arnaudon, et al.
0

It is well-known that irreversible MCMC algorithms converge faster to their stationary distributions than reversible ones. Using the special geometric structure of Lie groups G and dissipation fields compatible with the symplectic structure, we construct an irreversible HMC-like MCMC algorithm on G, where we first update the momentum by solving an OU process on the corresponding Lie algebra g, and then approximate the Hamiltonian system on G × g with a reversible symplectic integrator followed by a Metropolis-Hastings correction step. In particular, when the OU process is simulated over sufficiently long times, we recover HMC as a special case. We illustrate this algorithm numerically using the example G = SO(3).

READ FULL TEXT

page 1

page 2

page 3

page 4

research
07/07/2018

Note on the convergence time of some non-reversible Markov chain Monte Carlo methods

Introduced in statistical physics, non-reversible Markov chain Monte Car...
research
09/29/2018

Truly Concurrent Process Algebra Is Reversible

Based on our previous process algebra for concurrency APTC, we prove tha...
research
07/16/2020

Incorporating compositional heterogeneity into Lie Markov models for phylogenetic inference

Phylogenetics uses alignments of molecular sequence data to learn about ...
research
01/27/2020

Variational Optimization on Lie Groups, with Examples of Leading (Generalized) Eigenvalue Problems

The article considers smooth optimization of functions on Lie groups. By...
research
07/08/2021

Graded Symmetry Groups: Plane and Simple

The symmetries described by Pin groups are the result of combining a fin...
research
05/12/2020

Non-reversible guided Metropolis-Hastings kernel

We construct a non-reversible Metropolis-Hastings kernel as a multivaria...
research
08/15/2018

A novel Empirical Bayes with Reversible Jump Markov Chain in User-Movie Recommendation system

In this article we select the unknown dimension of the feature by re- ve...

Please sign up or login with your details

Forgot password? Click here to reset