Irreversible Langevin MCMC on Lie Groups

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 $\mathcal G$ and dissipation fields compatible with the symplectic structure, we construct an irreversible HMC-like MCMC algorithm on $\mathcal G$, where we first update the momentum by solving an OU process on the corresponding Lie algebra $\mathfrak g$, and then approximate the Hamiltonian system on $\mathcal G \times \mathfrak g$ 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 $\mathcal G = SO(3)$.

[1]  Radford M. Neal MCMC Using Hamiltonian Dynamics , 2011, 1206.1901.

[2]  K. Spiliopoulos,et al.  Irreversible Langevin samplers and variance reduction: a large deviations approach , 2014, 1404.0105.

[3]  Darryl D. Holm,et al.  Momentum Maps and Stochastic Clebsch Action Principles , 2016, 1604.04554.

[4]  Tudor S. Ratiu,et al.  Stochastic Variational Principles for Dissipative Equations with Advected Quantities , 2015, Journal of Nonlinear Science.

[5]  M. A. Clark,et al.  Accelerating staggered-fermion dynamics with the rational hybrid Monte Carlo algorithm , 2007 .

[6]  N. Pillai,et al.  A Function Space HMC Algorithm With Second Order Langevin Diffusion Limit , 2013, 1308.0543.

[7]  Darryl D. Holm,et al.  Noise and Dissipation on Coadjoint Orbits , 2016, Journal of Nonlinear Science.

[8]  A. Kennedy,et al.  Hybrid Monte Carlo , 1988 .

[9]  G. Pavliotis,et al.  Variance Reduction Using Nonreversible Langevin Samplers , 2015, Journal of statistical physics.

[10]  Alexis Arnaudon,et al.  Networks of coadjoint orbits: From geometric to statistical mechanics , 2018, Journal of Geometric Mechanics.

[11]  M. A. Clark,et al.  Shadow Hamiltonians, Poisson brackets, and gauge theories , 2013 .

[12]  Mark A. Girolami,et al.  Geometry and Dynamics for Markov Chain Monte Carlo , 2017, ArXiv.

[13]  Hamiltonian Monte Carlo on Symmetric and Homogeneous Spaces via Symplectic Reduction , 2019, 1903.02699.

[14]  E. Vanden-Eijnden,et al.  A patch that imparts unconditional stability to certain explicit integrators for SDEs , 2010, 1010.4278.

[15]  M. A. Clark,et al.  Tuning HMC using Poisson Brackets , 2008 .

[16]  Michela Ottobre,et al.  Markov Chain Monte Carlo and Irreversibility , 2016 .

[17]  Alessandro Barp,et al.  Hamiltonian Monte Carlo on Lie Groups and Constrained Mechanics on Homogeneous Manifolds , 2019, GSI.

[18]  E. Vanden-Eijnden,et al.  Pathwise accuracy and ergodicity of metropolized integrators for SDEs , 2009, 0905.4218.

[19]  Classical mechanics on group manifolds and applications to hybrid Monte Carlo , 1989 .

[20]  M. Betancourt,et al.  The Geometric Foundations of Hamiltonian Monte Carlo , 2014, 1410.5110.

[21]  T. Lelièvre,et al.  Free Energy Computations: A Mathematical Perspective , 2010 .