Global stability of a class of discrete-time recurrent neural networks

Sanqing Hu1, Jun Wang1
1Department of Automation and Computer Aided Engineering, Chinese University of Hong Kong, Hong Kong, China

Tóm tắt

This paper presents several analytical results on global asymptotic stability (GAS) and global exponential stability (GES) for the equilibrium states of a general class of discrete-time recurrent neural networks (DTRNNS) with asymmetric connection weight matrices and globally Lipschitz continuous and monotone nondecreasing activation functions. A necessary and sufficient condition is formulated to guarantee the existence and uniqueness of equilibria of such DTRNNS. The obtained results are less restrictive, different from, and improve upon the existing ones on GAS and GES of neural networks with special classes of activation functions.

Từ khóa

#Recurrent neural networks #Neural networks #Asymptotic stability #Lyapunov method #Symmetric matrices #Stability analysis #Sufficient conditions #Signal processing #Stability criteria #Design optimization

Tài liệu tham khảo

khalil, 1992, Nonlinear Systems michel, 1977, Qualitative Analysis of Large Scale Dynamical Systems 10.1109/72.914529 10.1109/9.917666 passino, 1998, Stability Analysis of Discrete Event Systems 10.1017/CBO9780511810817 10.1109/81.668873 10.1109/81.873882 10.1109/81.668877 10.1109/ICNN.1993.298717 jin, 0, equilibrium stability of discrete-time dynamical neural model, Proc 1993 World Congr Neural Networks 1993, 4, 276 10.1109/72.329692 10.1109/72.508944 10.1016/0893-6080(92)90011-7 10.1109/81.199861 10.1109/72.134288 10.1109/81.298364 10.1109/72.501720 10.1109/72.207617 10.1109/31.68298 10.1109/10.52325 10.1109/81.401145 10.1109/72.883412 10.1109/81.404065 10.1137/0614036 10.1073/pnas.81.10.3088 10.1109/81.641813 10.1073/pnas.79.8.2554 10.1109/72.182701 10.1109/IJCNN.1992.227277 10.1109/72.105420 10.1109/72.809094 10.1109/72.80289 10.1109/72.728385 10.1109/72.508941