摘要
文中主要提出了一种由模糊逻辑单元组成的聚类神经元网络,它可以用于数据的聚类分析.由于逻辑单元易于硬件实现和快速运算,因此大大提高了聚类分析的速度。尽管采用了竞争学习作为网络的学习算法,但是它却克服了一般竞争学习算法所固有的死点(deadunit)问题,使得聚类分析中初始点的选取可以有更大的随意性,这在具体应用中是很有用的.文中列举了一些有典型意义的实验,充分证实了这种方法的有效性.
A clustering neural network consisting of logic neurons is presented,which can be used indata clustering. Because logic operations are fast and easy to complete with hard ware,theywill increase the clustering speed. Although the network utilizes competitive learning as thelearning algorithm, it overcomes the probiem of a dead unit,so that there is more freedom inselecting the initial values,which is useful in practical applications.Results from the typicalexperiments illustrate the power and efficiency of the new method.
出处
《西安电子科技大学学报》
EI
CAS
CSCD
北大核心
1995年第1期1-7,共7页
Journal of Xidian University