On the Learning of Bayesian Network Graph Structures via Evolutionary Programming