A METHODOLOGY USING FUZZY-LOGIC TO OPTIMIZE FEEDFORWARD ARTIFICIAL NEURAL-NETWORK CONFIGURATIONS

After a problem has been formulated for solution by using artificial n eural network technology, the next step is to determine the appropriat e network configuration to be used in achieving a desired level of per formance. Due to the real world environment and implementation constra ints, different problems require different evaluation criteria such as : accuracy, training time, sensitivity, and the number of neurons used . Tradeoffs exist between these measures, and compromises are needed i n order to achieve an acceptable network design. This paper will prese nt a method using fuzzy logic techniques to adapt the current network configuration to one which is dose to (if not at) the optimal configur ation. The fuzzy logic provides a method of systematically changing th e network configuration while simultaneously considering all of the ev aluation criteria. The optimal configuration is determined by a cost f unction based on the evaluation criteria. The proposed methodology wil l be applied to an elementary classifier network as an illustration. T he procedure will then be used to automatically configure a network us ed to detect incipient faults in an induction motor as a real world ap plication.