A Generalization of the Noisy-MAX Parameterization for Biomedical Applications
Alessandro Magrini, Davide Luciani, Federico Mattia Stefanini · Florence Research (University of Florence) · 2016
The development of Bayesian networks for biomedical applications is often difficult because the number of parameters defining each conditional probability table grows exponentially with the increase in the number of parent variables. The Noisy-MAX parameterization have been extensively used to reduce the number of parameters defining a conditional probability table when causes independently influence the response. Unfortunately, the Noisy-MAX parameterization is not suited to a non-ordinal response variable. In this paper, we propose a generalization of the Noisy-MAX parameterization, called SoftDom parameterization, which is suited to a general biomedical response variable influenced by independent causal determinants.