Very large Bayesian multinets for text classification

Results of empirical evaluation of this BMN classifier are presented. The study suggests that tree-like Bayesian networks are able to handle a classification task in 100 000 variables with sufficient speed and accuracy.