| Summary: | 碩士 === 華梵大學 === 機電工程研究所 === 89 === The adjustment of the premise and consequent part of fuzzy if-then rules is the most important issue in fuzzy learning problem. This thesis presents a method combing self-organization and least square estimation to automatically adjust the parameters of a fuzzy system from training pattern. In chapter 2, we perform this method by using a feed-forward Tagagi-Sugeno-type fuzzy network on a typical plant of an inverted pendulum, and demonstrate the better convergent rate and average learning error when compared with some other traditional networks. In chapter 3, the proposed method is applied to a recurrent Tagagi-Sugeno-type fuzzy network, and the comparisons between the performance of using feed-forward and recurrent Tagagi-Sugeno-type fuzzy network are widely studied by an example of identification for a nonlinear system.
Finally, the merits and the drawbacks of the proposed hybrid method will be discussed for different kinds of learning objects, and we also cite the thought on the direction of application for the future.
|
|---|
