Learning Bayesian Networks

Abstract

Construction of a Bayesian belief network with domain experts is a di?cult and time-consuming task. In many domains are ill-understood. As a result, experts have problems in making causal relations explicit troubling the construction of network structures. It is this kind of domains where Bayesian belief networks are likely to be applied. In other domains where Bayesian belief networks are likely to be applied, there are few experts if any.Even if the causal relations in the domain can be made available easily, the de?nition of the assessment functions remains a time consuming task; human experts are good in judging qualitative relations but they are not good in quantizing these relations. Therefore, the build-test cycle may needtobe performed many times.