Martedì 21 settembre 2010, ore 12, Aula 101
Jim SMITH - Department of Statistics, University of Warwick
Introduce: M. Giovanna Ranalli, Università di Perugia
Abstract:

The Bayesian network is a common way of representing high dimensional discrete problems. Another is the event tree. In this talk I will demonstrate how certain classes of polynomial equations are intrinsic to both these classes of model and how understanding the geometry behind this semi-algebraic description enhances our understanding of the nature and implications of these model classes. The talk will be illustrated by two classes of model: binary BN tree models with hidden variables and event tree models with coalescence.