Synapse design based on memristor

摘要

the memristor is a passive two-terminal electrical device where its conductance is accurately modulated either by the charge or the flux flowing through it. In this paper, we implement a simple and flexible Verilog-A memristor model in order to perform synaptic functions, including both long-term potentiation LTP and long-term depression LTD, to subsequently validate the Hebbian learning algorithm (STDP). Our simulation results reveal the $I-V$ characteristic of a Ta2O5-based memristive device. We prove that the Verilog-A model is able to reproduce the conductance change in the STDP learning method. Therefore, the use of memristors as synapses in neuromorphic circuits may potentially provide high connectivity and high-density area for efficient computation.