Learning belief networks from data

Abstract
This paper presents an efficient algorithm for learning Bayesianbelief networks from databases. The algorithm takes a databaseas input and constructs the belief network structure as output.The construction process is based on the computation of mutualinformation of attribute pairs. Given a data set that is largeenough, this algorithm can generate a belief network very closeto the underlying model, and at the same time, enjoys the timecomplexity of O N ( )4on conditional independence...