EDML: A Method for Learning Parameters in Bayesian Networks

TL;DR

EDML is a novel method for learning MAP parameters in binary Bayesian networks with incomplete data, offering an alternative to EM and capable of learning maximum likelihood parameters with uninformative priors.

Contribution

The paper introduces EDML, a new parameter learning method for Bayesian networks, analyzing its properties and comparing it to EM.

Findings

EDML performs well compared to EM in various scenarios.

EDML can learn maximum likelihood parameters with uninformative Beta priors.

Theoretical and empirical analysis of EDML's properties.

Abstract

We propose a method called EDML for learning MAP parameters in binary Bayesian networks 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 compared to EM. We introduce EDML, explain its origin, and study some of its properties both analytically and empirically.