BPEXS: a learning rule for expert systems
S.E.T. Happee, R. Jäger, H.B. Verbruggen · 2003
The BPEXS learning algorithm for expert systems is presented. It adapts the confidence factors of a fuzzy reasoning expert system, using an algorithm based on the backpropagation learning rule for neural networks. The BPEXS algorithm was designed in a very general way. It can be applied to any expert system, provided that the product operator is chosen to implement the generalized modus ponens and the expert system's conclusions are defuzzified using the center-of-area method. Preliminary results show that the BPEXS algorithm can adapt the knowledge of an expert system to identify various second-order processes with reasonable accuracy.>