Approximation Strategies for Structure Learning in Bayesian Networks