IEEE Trans Neural Netw Learn Syst. 2021 Nov;32(11):5108-5117. doi: 10.1109/TNNLS.2020.3026912. Epub 2021 Oct 27.
This article investigates the stability of the switched neural networks (SNNs) with a time-varying delay. To effectively guarantee the stability of the considered system with unstable subsystems and reduce conservatism of the stability criteria, admissible edge-dependent average dwell time (AED-ADT) is first utilized to restrict switching signals for the continuous-time SNNs, and multiple Lyapunov-Kravosikii functionals (LKFs) combining relaxed integral inequalities are employed to develop two novel less-conservative stability conditions. Finally, the numeral examples clearly indicate that the proposed criteria can reduce conservatism and ensure the stability of continuous-time SNNs.
本文研究了时变时滞切换神经网络（SNN）的稳定性。为了有效保证具有不稳定子系统的系统的稳定性，并降低稳定性判据的保守性，本文首次利用允许的边依赖平均驻留时间（AED-ADT）来限制连续时间 SNN 的切换信号，并采用多个结合了松弛积分不等式的李雅普诺夫-克拉索斯基函数（LKFs）来提出两个新的不太保守的稳定性条件。最后，数值例子清楚地表明，所提出的准则可以降低保守性并确保连续时间 SNN 的稳定性。
