On the unconventional Hug integrator

Abstract: Hug is a recently proposed iterative mapping used to design efficient updates in Markov chain Monte Carlo (MCMC) methods when sampling along manifolds is of interest. In this paper we show that Hug may be interpreted as a consistent discretization of a system of differential equations with a rather complicated structure. The proof of convergence of this discretization includes a number of unusual features we explore fully. We uncover an unexpected and, yet, undocumented property of the solutions of the underlying dynamical system that manifest itself by the existence of Hug trajectories that fail to cover the manifold of interest. This suggests caution when using the Hug update.