| Summary: | 碩士 === 朝陽科技大學 === 資訊工程系碩士班 === 92 === In this thesis, a self-constructing fuzzy CMAC (FCMAC) model is proposed for various different applications. The Gaussian basis function is used to model the receptive field functions and the fuzzy weights for the FCMAC model. In order to make the better performance, we develop a self-constructing parametric fuzzy CMAC (called PFCMAC) model, based on the FCMAC model, which the Gaussian basis function is used to model the receptive field functions and the linear parametric equation of the model input variance is used to model the TSK-type outputs. Besides, if these two proposed model’s application domain is limited to static problems as a result of their internal feedforward network structure. To process temporal problems using these two proposed model are inefficient. Then the additional task is adopted which is the recurrent network is embedded in the FCMAC model by adding feedback connections with a receptive field cell to the FCMAC model, where the feedback units act as memory elements that can form the self-constructing recurrent fuzzy CMAC (called RFCMAC) model. An on-line learning algorithm is proposed to automatically construct the FCMAC model, the PFCMAC model, and the RFCMAC model, which consists of a structure learning scheme and a parameter learning scheme. The structure learning is based on the degree measure and the parameter learning is based on backpropagation algorithm. Finally, these three proposed models are applied in several simulations. Simulation results were conducted to illustrate the performance and applicability of these three proposed models.
|
|---|
