Learning recursive probability trees from probabilistic potentials

Abstract

A recursive probability tree (RPT) is an incipient data structure for representing the distributions in a probabilistic graphical model. RPTs capture most of the types of independencies found in a probability distribution. The explicit representation of these features using RPTs simplifies computations during inference. This paper describes a learning algorithm that builds a RPT from a probability distribution. Experiments prove that this algorithm generates a good approximation of the original distribution, thus making available all the advantages provided by RPTs