【摘要】 研究了带有时变时滞的递归神经网络的稳定性问题.假设该神经网络的神经元激励函数满足一般的扇形条件,通过使用Wirtinger不等式和倒凸组合法来估计Lvapunov-Krasovsii泛函的导数得到一个新的时滞相关稳定性判据.同时,应用凸包技术来处理时变时滞的导数,所提出的判据放松了时变时滞导数的限制.数值仿真结果验证了所得判据的有效性.更多还原

【Abstract】 The problem of stability analysis of recurrent neural network with time-varying delay was investigated with the neuron activation function being assumed to satisfy a general sector condition. By using the Wirtinger inequality and reciprocally convex approach to estimate the derivative of Lyapunov-Krasovsii functional, a new delay-dependent stability criteria were obtained. Moreover, the convex hull technique was applied to deal with the derivative of time-varying delay, the proposed criteria released the restriction on the derivative of time-varying delay. A numerical simulation has confirmed the effectiveness of the proposed criterion.更多还原