Szczuka M., Wróblewski J., 2006. A Rough-Neural Approach to Classifier Networks.

ABSTRACT

We discuss the notion of hierarchical concept (classifier) schemes. Processes of construction, tuning and learning of hierarchical structures of concepts (granules of knowledge) are presented. The proposed solution consists of a generalised structure of feedforward neural-like network approximating the intermediate concepts similarly to traditional neurocomputing approaches. Fundamental works are also supported by a more practical part where implementation and experimental verification of presented ideas is discussed. We provide the examples of compound concepts corresponding to the Bayesian and rule based classifiers, and show some intuition concerning their processing through the network.