Popis:
The lecture is based on the observation that a model for the statistical learning of genetic structure of populations due Dawson and Belkhir is based on the probabilistic notion of exchangeability. We can understand the model as defining probabilities in the space of partitions of discrete vector (genetic marker) data. The model has been extended, and applied, e.g., to problems of molecular ecology with good practical results. The implementation of the method is based on a a non-reversible Markov Chain Monte Carlo algorithm with parallell interacting searches, which has been shown to be convergent.