On learning mixed Bayesian networks
Aakanksha Bapna, G. Srinivasaraghavan · 2016
We propose a novel method for Bayesian learning of the parameters of a mixed belief network. Given the structure of a network, the parameters of conditional distribution of a node based on its type (discrete or continuous) and the types of its parents are learnt from the data. This node-wise updating scheme puts no restriction on the number and type of parents any node can have. We also extended the traditional algorithm for learning pure Gaussian networks to (i) deal with conditional Gaussian nodes, (ii) allow continuous nodes to be multivariate Gaussian and (iii) be able to converge to actual mean, covariance and weights of the network with which we generated the data.