机构地区: 西安交通大学电子与信息工程学院自动化科学与技术系
出 处: 《西安工业学院学报》 1997年第2期98-103,共6页
摘 要: 研究了小脑模型连接控制器(CMAC)的快速学习方法.首先分析了学习过程中学习干扰的原因及学习精度、学习次数、内存单元数之间的关系,然后基于内存单元的初始化和学习样本点的选择,构造了可快速精确地收敛于学习函数的快速学习方法———初始化随机法. A fast learning algorithm for CMAC(Cerebellar Model Articulation Controller) is investigated.The analysis of the learning interference by the generalization of CMAC and the relations among the training sessions, the associative cells and learning precision are presented. Based on the initialization of associative cells and the selection of the training samples, a fast learning algorithm for CMAC ——initial-random method, is proposed which can overcome effectively the learning interference, and can converge to the training function rapidly and precisely .The results shows that this method is effective.
领 域: [自动化与计算机技术] [自动化与计算机技术]