On the Validation of the DIMLP Neural Network
Guido Bologna · 2002
Rules are extracted from the DIMLP neural network in polynomial time with respect to the size of the clas-sification problem and the size of the network. With rules is possible to ask how well do inferences made compare with knowledge and heuristics of experts. Al-though fidelity of generated rules from the training set is 100%, perfect fidelity on new unknown data samples is not guaranteed. In this work we introduce a local dy-namic algorithm that makes rules consistent with new unknown cases. The presented method is computation-ally tractable and produces small changes in a rulebase.