摘要
联想神经网络的一个重要研究内容是收敛性。本文首先从联想神经网络构成算法的特点出发,用线性代数的方法对其演化特性进行了分析。在此基础上,对一类几何特征最为明显由投影学习算法构成的联想神经网络的演化过程进行了几何描述,并用几何的观点证明了这类网络在同步工作方式下的收敛性。这种分析方法能使我们对网络的演化特性有一更为直观、透彻的理解。
A key point of associative neural network is its convergence. From the point of view of the charactestic of its learning alsorithm, this paper, using linear alsebraic theory, analyses first its evolutionary characteristics, then a geometric description on the evolution of associative neural network with project learning algorithm is given and its convergence, under synchronous update mode, is proved using a geometric method, which can give a intuitive and good understanding of its evolution.
出处
《数据采集与处理》
CSCD
1992年第4期241-245,共5页
Journal of Data Acquisition and Processing
