Department of Mathematics, Chongqing Jiaotong University, Chongqing 400074, China.
School of Economics and Management, Chongqing Jiaotong University, Chongqing 400074, China.
Neural Netw. 2021 May;137:18-30. doi: 10.1016/j.neunet.2021.01.008. Epub 2021 Jan 20.
The problem on passive filter design for fractional-order quaternion-valued neural networks (FOQVNNs) with neutral delays and external disturbance is considered in this paper. Without separating the FOQVNNs into two complex-valued neural networks (CVNNs) or the FOQVNNs into four real-valued neural networks (RVNNs), by constructing Lyapunov-Krasovskii functional and using inequality technique, the delay-independent and delay-dependent sufficient conditions presented as linear matrix inequality (LMI) to confirm the augmented filtering dynamic system to be stable and passive with an expected dissipation are derived. One numerical example with simulations is furnished to pledge the feasibility for the obtained theory results.
本文研究了具有中立时滞和外部干扰的分数阶四元数值神经网络（FOQVNNs）的无源滤波器设计问题。本文通过构造 Lyapunov-Krasovskii 泛函和使用不等式技术，无需将 FOQVNN 分解为两个复值神经网络（CVNN）或 FOQVNN 分解为四个实值神经网络（RVNN），就推导出了时滞无关和时滞相关的充分条件，这些条件以线性矩阵不等式（LMI）的形式呈现，用于确定增广滤波动态系统的稳定性和被动性，并具有预期的耗散性。通过一个数值例子进行了仿真，证明了所得到的理论结果的可行性。
