Defining classes of influences for the acquisition of probability constraints for Bayesian networks
Linda C. van der Gaag, Eveline M. Helsper · European Conference on Artificial Intelligence · 2004
The task of eliciting all probabilities required for a Bayesian network can be supported by first acquiring qualitative constraints on the numerical quantities to be obtained. Building upon the concept of qualitative influence, we analyse such constraints and define a small number of influence classes. Based upon these classes, we present a method for efficiently acquiring the qualitative constraints that should be satisfied by the network's probabilities.