Stability And Synchronization Control Of Several Classes Of Neural Networks

As a mathematical model imitating the structure and function of human brain neurons,neural network can better simulate the working mechanism of the brain and plays an important role in artificial intelligence.With the continuous in-depth study of neural network,it has been widely and successfully applied in the fields of image recognition,optimization problem graph,associative memory and so on.These are based on the study of the dynamic behavior of neural networks.In addition,due to some external disturbances,physical limitations or uncontrollable factors,the dynamic behaviors of neural networks are not always ideal.Therefore,it is necessary and meaningful to exert external control on neural networks to make it have ideal dynamic behavior.This paper mainly focuses on the stability and synchronization control of several classes of neural networks.By using Halanay inequality and its generalized,the concept of Filippov solution,differential inclusion principle and some appropriate inequality scaling techniques,the stability and synchronization criteria of neural networks under corresponding control mechanisms are obtained.The main works of this paper are summarized as follows:The exponential input-to-state stability of a class of delayed neural networks with unmeasurable states under observer-based aperiodic intermittent control is studied.Combining the time-triggered control and the event-triggered control with aperiodic intermittent control respectively,the observer-based time-triggered aperiodic intermittent control and the observer-based event-triggered aperiodic intermittent control are designed.Then,by using the generalization of Halanay-type inequality,some sufficient conditions to ensure the exponential input-to-state stability of the system are obtained.In addition,the minimum activation time rate and control times of the observer-based event-triggered aperiodic intermittent control,the observer-based time-triggered aperiodic intermittent control and the observer-based periodic intermittent control are compared.The fixed-time synchronization of memristive Cohen-Grossberg neural networks with time delay and unmeasurable disturbance under sliding-mode control is discussed.By constructing an appropriate sliding-mode surface and using the concept of Filippov solution,the differential inclusion principle,the criterions for realizing the fixed-time synchronization of master-slave system are established.The asymptotic and finite-time cluster synchronization of complex-valued coupled neural networks with time-varying delays under linetype control are investigated.Based on Halanay inequality,the concept of Filippov solution and the differential inclusion principle,the criteria for realizing asymptotic and finite-time cluster synchronization of complex-valued coupled neural networks are obtained.In addition,the upper bound of the settling time for finite-time cluster synchronization is estimated.