ABSTRACT
The concept of developing computer-assisted decision aids was first envisioned shortly after the invention of the modern computer. In the last two decades, a number of such aids have been developed, initially using pattern recognition and simple neural network techniques. These were later supplanted for the most part by knowledge-based expert systems, utilizing either production rules or frames. Recently, neural network approaches have received new interest, due to theoretical advances as well as more powerful computers. It seems natural to develop expert systems which take advantage of the features of different technologies in the same system. The methodology described in this chapter utilizes rule-based techniques which encompass methods from approximate reasoning, and combines them with a neural network structure based on a non-statistical learning algorithm. The result is a system which can derive knowledge both from expert input and directly from accumulated data.
