School Seminars and Colloquia

Maybe some local algorithms aren't so bad...?

Statistical Mechanics/Combinatorics Seminar

by Tim Garoni


Institution: The University of Melbourne
Date: Mon 14th April 2008
Time: 1:00 PM
Location: Room 213, Richard Berry Building, The University of Melbourne

Abstract: The invention of the Swendsen-Wang "cluster" algorithm in 1987 has revolutionized the way people perform Monte Carlo simulations of many lattice models in statistical mechanics. A great deal of effort has been invested in devising generalizations of the Swendsen-Wang algorithm (which was originally applied to the Potts model) to simulate a number of different lattice models. Such cluster algorithms, which generically involve global updates of the
configuration, are often radically more efficient than traditional local algorithms near a critical point, and the standard wisdom today could be described as: "cluster algorithms are good, local algorithms are bad". In this talk I will discuss some quite simple local algorithms which give the cluster algorithms a run for their money.

For More Information: Iwan Jensen I.Jensen@ms.unimelb.edu.au