References & Citations
Statistics > Computation
Title: Geometrically Tempered Hamiltonian Monte Carlo
(Submitted on 4 Apr 2016 (v1), last revised 19 Feb 2017 (this version, v2))
Abstract: Hamiltonian Monte Carlo (HMC) has become routinely used for sampling from posterior distributions. Its extension Riemann manifold HMC (RMHMC) modifies the proposal kernel through distortion of local distances by a Riemannian metric. The performance depends critically on the choice of metric, with the Fisher information providing the standard choice. In this article, we propose a new class of metrics aimed at improving HMC's performance on multi-modal target distributions. We refer to the proposed approach as geometrically tempered HMC (GTHMC) due to its connection to other tempering methods. We establish a geometric theory behind RMHMC to motivate GTHMC and characterize its theoretical properties. Moreover, we develop a novel variable step size integrator for simulating Hamiltonian dynamics to improve on the usual St\"{o}rmer-Verlet integrator which suffers from numerical instability in GTHMC settings. We illustrate GTHMC through simulations, demonstrating generality and substantial gains over standard HMC implementations in terms of effective sample sizes.
Submission history
From: Akihiko Nishimura [view email][v1] Mon, 4 Apr 2016 14:24:38 GMT (239kb,D)
[v2] Sun, 19 Feb 2017 02:58:34 GMT (929kb,D)
Link back to: arXiv, form interface, contact.