摘要
首先推导了ＦｕｚｚｙＣ－Ｍｅａｎｓ算法在特征空间的迭代公式，然后就其隶属度信息在特征空间的分布缺陷提出两种改进方法：一是通过引入选择注意性参数控制隶属度信息的分布；二是从条件概率出发构造类置信度取代原隶属度．实验表明这两种方法均起到了较好的效果．
The iterative formulas of Fuzzy C Means algorithm in feature space were deduced before two methods were presented to improve the distribution property of membership information. The first was to control the distribution of membership information by introducting selective attention parameters. The other was to construct cluster fiducial values from the point of probability to replace original membership information. Results of the experiments confirm the effectiveness of the two methods.
出处
《红外与毫米波学报》
SCIE
EI
CAS
CSCD
北大核心
1999年第1期67-72,共6页
Journal of Infrared and Millimeter Waves
关键词
FUZZY
隶属度
选择注意性参数
置信度
FCM算法
fuzzy C means algorithm
membership values
selective attention parameters
fiducial values.
