Optimising and adapting the Metropolis algorithm
Seminar Room 1, Newton Institute
The Metropolis algorithm is a very popular method of approximately sampling from complicated probability distributions, especially those arising in Bayesian inference. A wide variety of proposal distributions are available, and it can be diĘ cult to choose among them. We will discuss optimal proposals under various circumstances. We will also consider the possibility of having the computer automatically "adapt" the algorithm while it runs, to improve and tune on the fly.