Learning Chordal Markov Networks by Dynamic Programming

Neural Information Processing Systems 

We present an algorithm for finding a chordal Markov network that maximizes any given decomposable scoring function.