Synchronization of CVNNs: A Time-Scale Impulsive Strategy

Qiuyuan Chen,Zhenkun Huang,Honghua Bin

doi:10.1109/access.2021.3059772

Qiuyuan Chen, Zhenkun Huang + Show 1 more

Open Access

https://doi.org/10.1109/access.2021.3059772

Copy DOI

Abstract

In this paper, we mainly investigate global exponential synchronization for master-slave complex-valued neural networks (CVNNs) under a time-scale impulsive strategy. CVNNs are separated into real and imaginary parts, which lead to two real-valued neural networks (RVNNs). Firstly, impulsive Halanay differential inequality on time scales as well as the comparison between general exponential function and exponential function in timescale sense is given based on the calculus of time scales. Then by constructing the appropriate Lyapunov functional and using the established lemma and proposition, the concepts of average impulsive interval (AII) and average impulsive gain (AIG), some novel synchronization criteria for the given master-slave CVNNs in impulsive form are obtained. Additionally, the convergence rate is estimated explicitly. Finally, one numerical example is given to show the effectiveness of the proposed results.

Highlights

As is well known, complex-valued neural networks (CVNNs) are more complicated than real-valued neural networks (RVNNs) because their states, connection weights and activation functions are all complex values
As is well known, CVNNs are more complicated than RVNNs because their states, connection weights and activation functions are all complex values
Many practical applications are obtained based on the properties of CVNNs like pattern processing [1], information science [2], and communication network [3]. [3] presented a novel auto-encoder by using complex-valued convolutional neural network to the modeling end-to-end MIMO wireless network, the method was both reasonable system performance and complexity reduction compared to the ML detector, especially in the field of drones

Summary

Introduction

CVNNs are more complicated than RVNNs because their states, connection weights and activation functions are all complex values. Many practical applications are obtained based on the properties of CVNNs like pattern processing [1], information science [2], and communication network [3]. The study of CVNNs becomes a hot spot, some results about dynamical analysis of CVNNs are emerged in [1]–[21]. The event-triggered exponential synchronization problem for a class of complexvalued memristive neural networks with time-varying delays was considered in [4]. The authors in [6] studied the issue of robust stability for a class of uncertain complexvalued stochastic neural networks with time-varying delays

Objectives

Results

Conclusion