Learning Bayesian Networks: A Unification for Discrete and Gaussian Domains

TL;DR

This paper introduces a unified Bayesian scoring metric for learning Bayesian networks applicable to both discrete and Gaussian domains, integrating prior knowledge with statistical data.

Contribution

It unifies previous methods for discrete and Gaussian Bayesian network learning by deriving a general Bayesian scoring metric applicable to both.

Findings

Derived a general Bayesian scoring metric for both domains

Unified discrete and Gaussian Bayesian network learning approaches

Utilized Dirichlet and normal-Wishart distributions for metric derivation

Abstract

We examine Bayesian methods for learning Bayesian networks from a combination of prior knowledge and statistical data. In particular, we unify the approaches we presented at last year's conference for discrete and Gaussian domains. We derive a general Bayesian scoring metric, appropriate for both domains. We then use this metric in combination with well-known statistical facts about the Dirichlet and normal--Wishart distributions to derive our metrics for discrete and Gaussian domains.