摘要
神经网络的稳定性被认为是神经网络各种应用的基础 主要利用网络的状态转移方程和能量函数来研究带有延迟项的离散Hopfield神经网络动力学行为 给出了延迟离散Hopfield神经网络收敛于周期小于等于 2的极限环的一些充分条件 给出了延迟网络收敛于周期为 2和 4的特殊极限环的一些充分条件 同时 ,得到了网络不存在任何稳定点的一些必要条件 所获结果不仅推广了一些已有的结论 。
The stability of neural networks is known to be basis of the network's various applications In this paper, the dynamic behavior of discrete Hopfield neural networks with delay is mainly studied by using the state transition equation and the energy function The sufficient conditions, under which the networks are convergent to cycle of length of 1 or 2, are present The sufficient conditions for the networks being convergent to special cycle of length of 2 and 4 are given Also, the conditions for the networks having no any stable states are achieved The obtained results not only generalize some existing results, but also provide a theoretical foundation of performance analysis and new applications of the networks
出处
《计算机研究与发展》
EI
CSCD
北大核心
2003年第4期550-555,共6页
Journal of Computer Research and Development
基金
国家自然科学基金 (699710 18)
