Modeling And Analysis Of Associative Memory Neural Network Based On Memristo

Memristor is a new type of circuit component,which can be used to build neural network circuits.Compared with the traditional CMOS transistor neural network circuit,the memristor neural network circuit has very strong computing power.In addition,memristor has the characteristics of small size,low energy consumption,high integration,fast reading and writing speed,so memristor is considered to be the best choice for building neural network circuits.In this paper,different neural network circuits are designed based on memristor,and the associative memory function of each neural network circuit is theoretically analyzed and simulated,so as to simulate the associative memory process in the biological brain.The main work of this article includes the following aspects:1.Three kinds of memristor models,namely HP memristor,threshold adaptive memristor and spin memristor,are studied through formula derivation and simulation experiments,and the advantages and disadvantages of the three kinds of memristor models are compared.After comprehensive analysis,it is determined to use HP memristor model to build neural network circuit.2.A memristor synaptic circuit composed of MOS transistors and HP memristor is proposed.Based on the memristor synaptic circuit,a neural network circuit simulating the associative memory function in Pavlov’s experiment is designed.The simulation experiment proves that the designed neural network circuit can imitate the associative learning,food forgetting and bell forgetting process in Pavlov’s experiment.3.Expand the memory resistive synaptic circuit in the previous chapter and design a neuron circuit.Based on the analysis and research of the working mechanism of associative memory,a neural network circuit for red rose feature recognition and red rose feature recall was designed.Simulation experiments have demonstrated that the designed neural network circuit can achieve corresponding feature recognition and recall functions.4.Based on HP memristor,a memristor bridge synaptic circuit is designed.On the basis of this synaptic circuit,a neuron circuit is formed by adding activation function circuit.Furthermore,a learning and recall associative memory neural network circuit was designed based on synaptic circuits and neural circuits.Finally,three different application experiments are designed: learning and recall of binary image,gray image and emotional image.The experimental results show that the designed neural network circuit can achieve learning and recall functions,and can resist certain external interference.To sum up,based on the HP memristor model,this paper designs three neural network circuits that can realize different associative memory functions,specifically studies the learning and forgetting process of dogs in Pavlov’s experiment,the feature recognition and feature recall process of red roses,and the learning and recall process of a variety of different images.Through theoretical analysis and simulation experiments,it has been proven that the designed neural network circuit can learn and associate memory like the biological brain.This article provides a reference for the further development of memory resistive neural networks in associative memory function.