Author(s)

Yanchun Liu, Shaotian Wang

Corresponding Author

Yanchun Liu

Abstract

Traditional electronic devices face challenges when dealing with large-scale and complex neural networks. As a nonlinear resistive device with memory function, memristor's resistance variability and memory characteristics provide new possibilities for the application of neural network. In this paper, an information processing circuit based on memristor is designed and implemented. The synaptic connection in neural network is simulated by memristor array, and the fast processing and response of input information are realized by combining with control circuit. In addition, an innovative information processing method is proposed, which uses the unique characteristics of memristor and the computing power of neural network to realize efficient and accurate information processing by constructing specific algorithm flow and optimization strategy. The experimental results show that the circuit design scheme can stabilize the weight update of analog neural network, and significantly improve the processing speed and accuracy compared with traditional methods when processing image, text and audio information. This study provides new ideas and technical support for developing an efficient and low-power neural network information processing system.

Keywords

neural network; Information processing circuit; memristor; information processing method