Cluster Monte Carlo algorithms and their applications

TL;DR

This paper reviews cluster Monte Carlo algorithms, highlighting their efficiency near critical points and their diverse applications, including in image processing, emphasizing their global update mechanism over local methods.

Contribution

It provides a comprehensive overview of cluster algorithms, their development, and applications across physics and other fields, emphasizing their advantages over local algorithms.

Findings

Cluster algorithms are highly efficient near critical points.

They enable large, global updates in Monte Carlo simulations.

Applications extend beyond physics to imaging processing.

Abstract

We review the background of the cluster algorithms in Monte Carlo simulation of statistical physics problems. One of the first such successful algorithm was developed by Swendsen and Wang eight years ago. In contrast to the local algorithms, cluster algorithms update dynamical variables in a global fashion. Therefore, large changes are made in a single step. The method is very efficient, especially near the critical point of a second-order phase transition. Studies of various statistical mechanics models and some generalizations of the algorithm will be briefly reviewed. We mention applications in other fields, especially in imaging processing.