Mixed H∞/passive synchronization for persistent dwell-time switched neural networks via an activation function dividing method

• Published in: Applied mathematics and computation 2023-04, Vol.442, p.127718, Article 127718
• Main Authors: Wang, Xiaomin, Li, Feng, Hu, Xingliu, Wang, Jing
• Format: Article
• Language: English
• Subjects: [formula omitted]/passive synchronization; Activation function dividing method; Persistent dwell-time switching law; Switched neural networks
• ISSN: 0096-3003; 1873-5649
• DOI: 10.1016/j.amc.2022.127718

•In order to describe the switching between parameters in NNs, a more general PDT switching law is introduced in this paper, which makes the considered neural network model more realistic.•When dealing with the synchronous control problem of NNs, the mixed performance used in this paper is more general than the individual H∞ or passive performance.•In this paper, AFDM is used to study the synchronization of PDT switching NNs, where the activation function is divided into two sub-intervals to reduce the conservatism of the system stability criterion. The mixed H∞/passive synchronization issue of discrete-time switched neural networks is studied in this paper. In order to regulate the switching between subsystems, the persistent dwell-time switching law is adopted. The paper aims to design a suitable synchronization controller to make the synchronization error system satisfy the mixed H∞/passive performance and achieve global uniform exponential stability. By employing Lyapunov stability theory, performance analysis criteria and the synchronization controller design method are given, in which an activation function dividing method is employed to reduce their conservatism. Simulation results demonstrate the superiority and effectiveness of the method.