Learning with privileged information using Bayesian networks