Condensed knowledge representation in BP-networks

In the framework of NN-theory, a lot of research deals with designing self-organizing neural networks with an internal structure that seems to be appropriate for a particular task domain. The aim of this paper is to contribute to better understanding the behaviour of BP-networks, their knowledge extraction and generalization capabilities. This is the way along which neural networks and rule-based AI-systems are generally hoped to unify. The author proposes an algorithm for adjusting weights in layered networks in order to create a condensed internal representation. Experimental results are briefly referred to