Finite-Time and Fixed-Time Stabilization of Discontinuous Neural Networks With Nonlinear Coupling via Switching Approach

Chao Yang,Yicheng Liu,Lihong Huang

doi:10.1109/access.2020.3025458

Abstract

This brief discusses the finite-time and fixed-time stabilization problem of nonsmooth switched neural networks with nonlinear coupling and delays. Base on the framework of Filippov theory, methodology of graph theory and classical Lyapunov functional technique, one can optimize both the finite-time stabilization and fixed-time stabilization issues for complex nonlinear model by utilizing two novel state-feedback controllers and the corresponding adaptive controllers. Moreover, how to shorten the stabilization settling time is discussed in detail when the complex neural network systems are strongly connected, and the influence of initial value on the system is deeply analyzed. Finally, two examples are given to show the advantages and experimental analysis of this method.

Highlights

Neural network system is based on computer simulation of brain development network, which is one of the ’artificial intelligence’ field technologies
We introduce a class of complex neural network with nonlinear coupling and discontinuous activations
Based on the above discussion and a large number of computations, this paper investigates a class of complex delayed neural network with nonlinear coupling and discontinuous activation, our goal is to improve system stability in finite time and fixed time

Summary

Introduction

Neural network system is based on computer simulation of brain development network, which is one of the ’artificial intelligence’ field technologies. It has occupied a more dominant position and aroused wide attention of scholars. The neural network system with discontinuous characteristics is one of the most concerned systems, which has been studied in many literatures [1]–[7]. The information transmission between different neurons is generally discontinuous, which makes the neural network system more complex and closer to the real experimental environment. In order to accurately describe the experimental results, it is important to study the discontinuous neural networks. In [8], Wang and his

Objectives

Results

Conclusion