Research And Design Of Neural Network Chaotic Dynamics Based On Memristor

Memory is an important characteristic of biological neurons,and memristor is an electronic component with memory characteristics.Scholars have proved that memristor is a natural material to simulate ion channels and neural networks of neurons.Memristor neural network model emphasizes the memory effect of biological neural system by using memristor.Compared with general neural network system,memristor neural network is closer to the actual neural network system of biological body,the dynamic characteristics of memristor neural network can better reflect the specific functions of actual biological neural network.There are chaotic dynamic behaviors in the brain of organisms,and chaotic behaviors are closely related to the cognitive function,information recognition function and memory function of organisms.Therefore,chaotic dynamics of neural network based on memristor has important research significance and value.Based on the study of the literature on chaotic dynamics of memristor neural network,a new model of local active memristor is proposed,the dynamic characteristics of local active memristor neural network are discussed,and the dynamic characteristics of memristor electromagnetic radiation neural network are studied(1)A new model of local active memristor is proposed,and its equivalent circuit is built by using analog devices.A new local active memristor neural network system is constructed by using the proposed local active memristor,and the dynamic characteristics of the local active memristor neural network system are explored.It is found that the system exhibits complex chaotic dynamics in the local active region and stable resting behavior in the local passive region;There is a new dynamic phenomenon in the system,which is the behavior of coexistence position symmetry;With the change of the local active memristor strength,the local active memristor neural network system presents a variety of dynamic alternating phenomena.Finally,the hardware equivalent circuit verifies the effectiveness of the numerical simulation.(2)In this paper,a neural network model without electromagnetic radiation is proposed,and its dynamic characteristics are explored by using the numerical analysis methods such as Li's index and phase diagram.It is found that the system only has a single periodic dynamic behavior.On this basis,two different memristor models are used to describe the influence of electromagnetic radiation on neurons,and a new memristor electromagnetic radiation neural network system is proposed.The dynamic characteristics of the memristor electromagnetic radiation through the neural network system are explored.It is found that with the change of electromagnetic radiation intensity,the system appears rich dynamic behavior,under different initial conditions,the system appears multi stable phenomena including hidden hyperchaotic dynamics,which has not been found in HR neural network before.Finally,the effectiveness of numerical simulation is verified by hardware circuit experiment.