RRAM-Based Binary Neural Networks Using Back-Propagation Learning

摘要

Hardware binary neural networks (BNNs) based on resistive random access memory (RRAM) are designed and investigated in this work. RRAM devices that work in binary mode are used as electronic synapses. The simulation results indicate that the designed BNNs can achieve an accuracy of 94% on the MNIST database, and show remarkable tolerance to non-ideal properties of RRAM-based electronic synapses.