Combining fuzzy pattern recognition and fuzzy control in an AI driven neural network

We combined fuzzy pattern recognition and fuzzy control in the testing program of a fuzzy ANN. In pattern recognition for each input vector the network provides a spectrum of "object to a class belongingness" with respect to all classes. A fuzzy controller using some of the components of the same input vector generates a numerical output by simulating an equation. The choice of these components and selection of the equation is carried out by a rule-based expert system, the input of which is the spectrum generated by the fuzzy pattern recognition network. A technique that reduces the training times of fuzzy ANNs is also described. This hybrid AI/ANN system is intended for applications where the pattern recognition signals can be used for control. One such application is discussed.