An Isaac Newton Institute Workshop

Recent Advances in Monte Carlo Based Inference

Applications of Extended Ensemble Monte Carlo

1st November 2006

Author: Yukito Iba (The Institute of Statistical Mathematics)

Abstract

"Extended Ensemble Monte Carlo" is proposed as a generic term which indicates methods such as Parallel Tempering and Multicanonical Monte Carlo. These methods sample a composition or extension of given distributions, and provide a simple and practical way to attack slow mixing and ergodicity problems. The purpose of the talk is not to present a novel algorithm, but explore variations and applications of the algorithms, including their use in statistical physics, combinatorics, and rare event sampling. It is stressed that overcoming slow mixing is a key for the extension of application fields of MCMC.