A characterization of the Dirichlet distribution through global and local parameter independence

Abstract
This article is a characterization ofthe Dirichlet distribution based on local and global parameter independence and on theassumption that the probability distributions of all the parameters are positive pdfs. InSection 3 we explore the circumstances under which our characterization implies that thedistribution of the parameters associated with each node in a Bayesian network must beDirichlet, in which case, the fifth assumption for learning is actually redundant. The as3sumption of...

This publication has 0 references indexed in Scilit: