Learning Bayesian networks by hill climbing: efficient methods based on progressive restriction of the neighborhood | Publicación