Seol mar théacs é seo: Approximation methods for efficient learning of Bayesian networks