Case Studies
Summer 2009
Jonathan Mascie-Taylor
Multi-level coarse graining methods for sampling stochastic particle systems.(report)
We considered d dimensional Ising type models and apply coarse graining and parallel tempering techniques. We studied these techniques and combined them to create a new method which had increased convergence to equilibrium and an improved ability to explore multimodal distributions. The method worked by creating independent replicas of the model being simulated at dijlent temperatures and levels of resolution; and then exchanging information between them. Whilst this new combined technique was applied to the Ising model it has the potential to be applied to many other areas of science. We examined dijlent ways of measuring the rate of convergence and used these measures to compare the new method with traditional methods.