Simulation Analysis Of Memristor-based Spiking Neural Network

With the continuous development of science and technology,the electromagnetic environment is becoming more and more complicated.The system with the traditional protection methods shows its shortcomings in coping with the complicated and changeable electromagnetic environment,and the biology system shows the adaptive,immunity and self-repairing ability in complex electromagnetic environment has attracted wide attention,the concept of electromagnetic bionic protection came into being.As the third generation artificial neural network,spiking neural networks is considered as the model closest to the real biological neuron signal transmission and network characteristics.It is the key to realization of neuromorphic computing and applied to the electromagnetic bionic protection.Spiking neural networks encodes real-world information into pulse time,which makes it able to decode information quickly.It is widely used in pattern recognition,image processing,associative memory,etc.In electromagnetic bionic protection,memristor(memory resistor),a new bionic device,have memory,nanoscale,low power consumption and many other features,can be used as a bionic synaptic into the spiking neural network.This paper investigated the construction of memristor-based spiking neural network for bionic hardware system,from the demand of electromagnetic bionic protection,based on the spiking neuron Izhikevich(I).The memristor was used as the weight for the Izhikevich neuron to construct the memristor-based spiking neural network.The main works of this paper are as follows:(1)The mathematical model and electric properties of the memristor were analyzed.Based on the nonlinear boundary migration model,the model was constructed by the tool of Simulink.The current-voltage,flux-charge,memristor value-voltage of memristor model were analyzed under different input signals.(2)The mathematical model of Izhikevich(I)was analyzed,and the model was constructed with the tool of Simulink.The simulation analysis of Izhikevich(I)model was completed under five different discharge patterns,Regular Spiking,Fast Spiking,Intrinsically Burst,Chattering,Fast Spiking and Low-threshold Spiking.(3)The mathematical model of memristor and the model of spiking neuron Izhikevich(I)were discretized by using Euler method,and the model of memristor and spiking neuron were constructed and simulated with the tool of DSP Builder.(4)Based on the characteristics of the memristor,the memristor was used as the weight input part of the neuron Izhikevich(I).Based on the topology structure of the artificial neural network,the model of memristor-based spiking neural network was constructed and Action potentials were compared and analyzed.