ABSTRACT
This chapter discusses how neural networks can be used for diagnostic expert systems. The specific problem presented is scoring site plans on an architecture certification examination. A hybrid system containing a rule-based expert system and a neural network for this problem is described. Techniques for knowledge engineering using neural networks to represent expert knowledge are presented. In particular, it is shown that the knowledge put into rules can also be put into neural networks, and that other knowledge which is hard to formulate as rules is easy to represent in networks.
