RRAM-based pattern recognition system with locally inhibited post-neurons

Abstract

A novel RRAM-based pattern recognition system with locally inhibited post-neurons is developed. The system is able to learn the whole MNIST training set (60,000 patterns). By using the system, the same post-neuron is fired by the similar patterns in the same training class, which causes the reduction of hardware cost. With the locally inhibited post-neuron, the system can achieve more than 90.73% recognition accuracy.