 |
 Bayesian and credal network
Bayesian networks are one of the most popular tools for reasoning in uncertain domains. By structuring the domain by means of cause-effect relationships, the model allows the power of probability theory to be fully exploited for applications. Bayesian networks enable predictive and diagnostic reasoning to be realized by a probability propagation in the graphical structure that represents the cause-effect relationships. Bayesian networks can be automatically inferred from the data alone and be used for the specific purpose of discovering knowledge in databases. Credal networks extend Bayesian networks in the direction of robustness. A credal network can be induced from a small or an incomplete database, and still guarantee that inferences are robust. |
|
