基于遗忘效应忆阻器的LIF神经元电路研究

LIF neuron circuit based on forgetting effect memristors

细胞神经网络(CNN)被公认为是一种强大的大规模并行网络架构,能够高速执行运算操作和解决复杂的工程问题,但是目前关于硬件实现神经元的研究处于起步阶段。首先,研究了一个基于SrTiO_(3)(STO)的忆阻仿真模型,并分析了该模型的阻值变化特性与磁滞回线。其次,在此基础上设计了基于忆阻器的LIF神经元电路,验证了忆阻器模型可很好地结合到该神经元电路中。最后,通过PSpice仿真实验分析了突触前神经元、突触权重以及输入信号频率对于膜电位的影响,验证了基于遗忘模型忆阻器构成的LIF神经元电路可实现对输入时间信息和空间信息的整体反应。

Cellular neural networks(CNNs)are generally considered to be a powerful massively parallel network architecture,which is capable of performing arithmetic operations at high speed and solving complex engineering problems.However,the current research on hardware implementation of neurons is in its infancy.In this work,firstly,a memristive simulation model based on SrTiO_(3)(STO)was investigated and the resistance variation characteristics and hysteresis lines of this model were analyzed.Secondly,a leaky-integrate-and-fire(LIF)neuronal circuit based on the memristor was designed based on this model,and it turned out that the memristor model can be well integrated into this neuronal circuit.Finally,through the PSpice simulation experiments,the effects of presynaptic neurons,synaptic weights and input signal frequency on the membrane potential were analyzed.At the same time,it was verified that the constitutive LIF neuronal circuit based on the forgetting-based model memristors could realize the overall response to the input temporal and spatial information.