Integrated Crossbar Array With Resistive Synapses and Oscillation Neurons
Integrated Crossbar Array With Resistive Synapses and Oscillation Neurons
复制标题
DOI：
10.1109/led.2019.2921656
复制
发表时间：
2019-08-01
影响因子：
4.9
通讯作者：
Yu, Shimeng
中科院分区：
文献类型：
--
作者：
Woo, Jiyong;Wang, Panni;Yu, Shimeng
在这封信中，我们制造了一个交叉杆阵列，其结构类似于神经网络的一列，其中一个神经元与多个突触并行连接，以进行芯片上集成。我们没有使用复杂的CMOS神经元电路，而是在交叉开关阵列的边缘集成了一个阈值开关作为一个紧凑的振荡神经元，它将加权和转换为一个振荡频率。当输入向量被加载到阵列的多行中时，振荡频率被测量为与模拟列电流成比例。这是具有突触和神经元的集成交叉开关阵列的第一个实验演示，为使用新兴的神经形态计算设备技术实现完全并行计算和处理铺平了道路。
In this letter, we fabricate a crossbar array that structurally resembles a column of the neural network, where one neuron is connected with multiple synapses in parallel for on-chip integration. Instead of using complex CMOS neuronal circuit, we integrate a threshold switch at the edge of the crossbar array as a compact oscillation neuron, which converts the weighted sum to an oscillation frequency. When the input vectors are loaded into multiple rows of the array, the oscillation frequency is measured to be proportional to the analog column current. This is the first experimental demonstration of an integrated crossbar array with both synapses and neurons, paving the path to a fully parallel computation and processing using emerging device technologies for the neuromorphic computing.