Parallel Tempering with Equi-Energy Moves

Meili Baragatti (1) , Agnès Grimaud (1) , Denys Pommeret (1)

1 IML - Institut de mathématiques de Luminy

The Equi-Energy Sampler (EES) introduced by Kou et al. [2006] is based on a population of chains which are updated by local moves and equi-energy jumps. This algorithm has been developed to facilitate global moves between the different chains, resulting in a good exploration of the states space by the target chain. This method seems to be more efficient than the classical Parallel Tempering (PT) algorithm. However it necessitates increased storage and the convergence of the original EES is not guaranteed (see Andrieu et al. [2008]). In this paper we propose an adaptation of the EES that combines PT with the principle of jumping between chains with same levels of energy. This adaptation, that we shall call Parallel Tempering with Equi-Energy Moves (PTEEM), keeps the original idea of the EES method and ensures convergence. Performances of the PTEEM algorithm are compared with those of the EES and of the standard PT algorithm in the context of mixture models.

Mots clés: population-based MCMC; Algorithm convergence; equi-energy sampler; mixture models; parallel tempering; population-based MCMC.

Domaines: Méthodologie [stat.ME]

Meili Baragatti, Agnès Grimaud, Denys Pommeret. Parallel Tempering with Equi-Energy Moves. 2011. ⟨hal-00559174v1⟩