Decision-Theoretic Methods for Learning Probabilistic Models