Iterative simulation methods (Q918185): Difference between revisions

For computing quantities related to a probability measure \(\pi\) which is difficult to handle directly, one way is to construct a Markov transition P, admitting \(\pi\) as unique invariant measure and to use the convergence \(\mu P^ n\to \pi\). These iterative simulations methods, often quoted as Metropolis' methods, are discussed here from the point of view of the choice of P, the speed of convergence and the efficiency of the implementation. Two families of examples are explained clearly which are relevant for these methods: Markov fields (Potts and Ising models) and point processes. The speed of convergence is, by general results, geometrical, but examples are given in which slow or surprising behaviours occur.