Robust learning Bayesian networks for prior belief
This work addresses a specific technical issue in Bayesian network learning for researchers and practitioners, representing an incremental improvement.
The paper tackles the sensitivity of learning Bayesian networks to the equivalent sample size (ESS) in the BDeu score, which causes unstable results, by analyzing the reasons for this sensitivity and proposing a robust learning score that eliminates sensitive factors from the log-BDeu approximation.
Recent reports have described that learning Bayesian networks are highly sensitive to the chosen equivalent sample size (ESS) in the Bayesian Dirichlet equivalence uniform (BDeu). This sensitivity often engenders some unstable or undesirable results. This paper describes some asymptotic analyses of BDeu to explain the reasons for the sensitivity and its effects. Furthermore, this paper presents a proposal for a robust learning score for ESS by eliminating the sensitive factors from the approximation of log-BDeu.