A New Bayesian Tree Learning Method with Reduced Time and Space Complexity

Abstract

Bayesian networks have many practical applications due to their capability to represent joint probability distribution in many variables in a compact way. There exist efficient reasoning methods for Bayesian networks. Many algorithms for learning Bayesian networks from empirical data have been developed. A well-known problem with Bayesian networks is the practical limitation for the number of variables for which a Bayesian network can be learned in reasonable time. A remarkable exception here is the Chow/Liu algorithm for learning tree-like Bayesian networks. However, its quadratic time and space complexity in the number of variables may prove also prohibitive for high dimensional data. The paper presents a novel algorithm overcoming this limitation for the tree-like class of Bayesian networks. The new algorithm space consumption grows linearly with the number of variables n while the execution time is proportional to n·ln(n), hence both are better than those of Chow/Liu algorithm. This opens new perspectives in construction of Bayesian networks from data containing tens of thousands and more variables, e.g. in automatic text categorization.