Method Research Of Neuronal Circuits Implementation Based On Memristor

Artificial neural network is an important method of modern information processingand intelligent control field. Realization of hardware neural network technology is animportant aspect of neural network research. From the definition of the neural networkcan know that neurons are the basis of neural networks. Traditional analog neuroncircuits can not accurate adjust and convenient storage weights. So, design of a highprecision, good stability, convenient adjustment artificial neuron circuit undoubtedly hasgreat practical value.Memristor is the fourth basic circuit element in addition to resistor, capacitor andinductor. If we adjust the voltage or current cross two-terminals of memristor, thememristance will be changed. Memristor is a memory function with nanoscaledimensions of nonlinear two-terminal passive devices.This paper presents a memristor-based artificial neural circuits, which can adjust thedirection and magnitude of the synaptic weights and achieve signal input by a port.Because of the resistance of memristor is continuously changing, and has a memoryfunction, therefore, The neuronal circuits designed in this paper, the weight adjustmentprecision, weights storage stability than traditional analog neuron circuit has a greatadvantage. The design in this paper is proved to be reasonable and feasible through aseries of SPICE simulations.