Research of GreyCMAC Learning and Its Applications

• Main Authors: Po-Lun Chang, 張博綸
• Other Authors: Ying-Kuei Yang
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/63738890706604904501

博士 === 國立臺灣科技大學 === 電機工程系 === 98 === The main purpose of this thesis is to improve the learning performance of CMAC (Cerebellar Model Articulation Controller). Using grey relational analysis based approach, adaptively regulate both learning rate and updating memory cells to increase the performance of CMAC. Additionally, some applications of CMAC are also discussed in the thesis. The advantages of supervised CMAC neural network are fast learning convergence, capable of mapping nonlinear functions quickly due to its local nature of weight updating, simple architecture, easily processing and hardware implementation. CMAC could be applied to intelligent control, pattern recognition, signal processing, data mining, robotics and other fields. In learning phase, the neighbor memory cells of CMAC have the phenomena of learning interference. The phenomena evidently reduce the learning performance of CMAC. Therefore, fast learning and accurate convergence are the two issues to be most concerned in the research area of CMAC. The difficulty of research in this field is to improve the trade-off between fast learning and accurate convergence over that of existing methods. In order to improve the learning speed and accuracy simultaneously, this thesis investigates to incorporate grey relational coefficients with number of training iterations to obtain an adaptive and appropriate grey learning rate for each input state to improve the CMAC stability and convergence. Additionally, this thesis also proposes that the amount of weight adjustment to a memory cell of an addressed memory cell must be relational to the trained input area, grey relational grade in the current training iteration and the inverse of the number of learning times to minimize the learning interference. A novel credit apportionment approach is thus derived for implementing this idea to achieve fast and accurate learning performance. The results of the experiments and various applications conducted in this study clearly demonstrate that the proposed approach provides a more accurate learning mechanism and faster convergence. The proposed CMAC model with fast learning and accurate convergence can perform adequately in various applications.