UC Irvine Today
On a Quasi-Bayes Procedure for Recursive Learning in Mixture Models
The Department of Statistics is proud to present Sonia Petrone, Bacconi University, Milan. Although originally proposed as an approximation of the Bayesian solution, its quasi-Bayes properties remain unclear. The speaker and team propose a novel methodological approach. They regard the algorithm as a probabilistic learning rule, that implicitly defines an underlying probabilistic model; and we find such model. They can prove that it is, asymptotically, a Bayesian, exchangeable mixture model, and show that, under some conditions, it implies a novel nonparametric prior on densities.
Thursday, October 10 at 4:00pm to 5:00pm
Donald Bren Hall, 6011
6210 Donald Bren Hall, Irvine, CA 92697