摘要
构造模糊神经网络时确定初始的隶属函数是一个难点,提出了一种新的基于粗糙集理论的隶属函数获取算法,该算法根据粗糙集理论中基于属性重要性的离散化方法确定条件属性的断点,再通过断点确定各模糊集合隶属函数的中心和宽度,同时给出了网络各参数的修正公式;仿真结果证明,该算法在学习的快速性和精度上具有良好的性能。
How to determine the membership function is a nodus while modeling the Fuzzy Neural Networks. A novel algorithm based on Rough Set theory for acquiring membership function is presented . The parameters of membership function are determined according to the segmentation dot which are computed by the method of data diseretization on the basis of the importance of attribute values. The modifying formulas of Fuzzy Neural Networks are also put forward. Finally , the simulation proves that the proposed method shows good performance in the learning speed and accuracy.
出处
《计算机测量与控制》
CSCD
2006年第6期782-784,789,共4页
Computer Measurement &Control
基金
国家自然科学基金资助项目(60474032)
关键词
粗糙集
模糊神经网络
隶属函数
离散化
rough set
fuzzy neural network
membership function
discretization
