EDML: A Method for Learning Parameters in Bayesian Networks

摘要

We propose a method called EDML for learn ing MAP parameters in binary Bayesian net works under incomplete data. The method assumes Beta priors and can be used to learn maximum likelihood parameters when the priors are uninformative. EDML exhibits interesting behaviors, especially when com pared to EM. We introduce EDML, explain its origin, and study some of its properties both analytically and empirically.