The Stability Analysis Of Variable-parameters Neural Networks

The paper has firstly systematically studied the stability of neural networks with variable coefficients, and mainly studied the four following problems:Chapter 2 has firstly sysematically studied L2-gain stability of Hopfield neural networks, and extended constant coefficients to variable coefficients, the conclusions extended index[4-9];Chapter 3 has studied .the existence and unique of equilibrium of neural networks with variable coefficients and also studied the stability of neural networks with variable coefficients and provided sufficient conditions and sufficient-efficient conditions of local stability and global stability when the unique equilibrium exists.The conclusions extended index[1417];. Chapter 4 has studied the velocity of neural networks tendency to equilibrium when the unique equilibrium exists, and gave sufficient-efficient conditions of exponential stability.The conclusions extended index[21-22].This paper has studied the four problems. According to the characteristics of neural networks models, making use of the liapunov function, combining with the method of inequality analysis and the method of variations of the parameters, the stability of neural networks with variable coefficents is discussed in depth. Some per-fect laws have been gained and others'inconclusions have been extended.