时滞递归神经网络全局稳定性的新判据

New criteria for global stability of recurrent neural networks with delays

借助于非光滑分析和Lyapunov函数法,研究了一类递归神经网络模型,并得到了几个确保延时RNNs系统存在全局渐近稳定的平衡点的充分条件.在这些结果中,既没有假设连结权矩阵是对称的,也没有要求激活函数是有界的、光滑可导的.最后,给出一个例子说明结论的有效性.

In this paper, the global asymptotic stability and the existence of equilibrium are considered for a general class of recurrent neural networks with delays by using nonsmooth analysis and Lyapunov methods. New criteria are obtained, which does not require the boundedness of activation functions and the symmetry of the connection matrices. An example is also given to illustrate the effectiveness of the results.