Learning a L 1-regularized Gaussian Bayesian network in the space of equivalence classes