Closed-Form Learning of Markov Networks from Dependency Networks

Daniel Lowd

Markov networks (MNs) are a powerful way to compactly represent a joint probability distribution, but most MN structure learning methods are very slow, due to the high cost of eval…