Improved generalization of neural classifiers with enforced internal representation

Abstrakt

Unambiguous condensed internal knowledge representation improves generalization capabilities of BP-networks.