基于增广泛函的混合时滞区间神经网络的鲁棒稳定性

本文针对一类带有混合时变时滞 (离散和分布时滞)的区间递归神经网络进行了全局鲁棒稳定性研究.与之前的处理方法不同,在本文中通过使用一种新型的增广Lyapunov-Krasovskii泛函,从而得到了一类新颖的关于区间递归神经网络的时滞依赖全局鲁棒稳定性判据.在新的增广泛函中,由于首次使用了带有激活函数的积分项,系统状态和激活函数之间的关系将被更好地表示出来.因此,本文提出的判据具有更小的保守性.同时,在本文提出的判据中,放松了时变时滞变化率必须小于1的限制.仿真结果进一步证明了本文结果的有效性.

Robust stability of interval neural networks with mixed time-delays via augmented functional

Global robust stability of interval recurrent neural networks with mixed time-varying delays (discrete timevarying delay and distributed time-varying delay) is investigated. Being different from existing reports, the novel delaydependent robust stability criteria for interval recurrent neural networks with mixed time-varying delays employ a new augmented Lyapunov-Krasovskii functional. In the new augmented functional, we introduce an integral term to the activation function, which gives a preferable representation of the relation between states of the system and the activationfunction. Because of the new functional, the criteria proposed in this paper are less conservative than the currently existing ones. Moreover, the employment of the Jensen inequality in proving the criteria relaxes the restriction on the time derivative of the time-varying delay in the proposed criteria. The simulation is provided to verify the effectiveness of theproposed results.