一类一般输入输出函数的离散神经元模型的分支  

The Bifurcation of a Class of Discrete-Time Neural Networks with a General Activation Function

作　　者：陈薇娜[1] 阮炯[1] 

机构地区：[1]复旦大学数学科学学院,上海200433

出　　处：《数学年刊（A辑）》2009年第6期793-802,共10页Chinese Annals of Mathematics

摘　　要：利用经典分支理论研究了一类一般输入输出函数的离散神经元模型的分支问题,得到了该类模型产生倍周期分支和鞍-结点分支的充分条件,推广了目前特殊的正弦输入输出函数的该类模型的结果.所得的结果为这一类神经网络的应用提供了重要的理论基础.By using classical bifurcation theories, the authors investigate a class of discretetime neural networks with a general activation function, and obtain the sufficient condition of period-doubling bifucation and saddle-node bifucation of this model, which can be regarded as an extension of a sinusoidal activation function. As a result, an important theoretical foundation for the application of this class of neural networks is provided.

关 键 词：离散神经元模型 倍周期分支 鞍-结点分支 

分 类 号：O175[理学—数学]