Neural Network Activation Functions with Electro-Optic Absorption Modulators
Neural Network Activation Functions with Electro-Optic Absorption Modulators
复制标题
电光吸收调制器的神经网络激活功能
DOI：
10.1109/icrc.2018.8638590
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
V. Sorger
中科院分区：
文献类型：
--
作者：
J. George;A. Mehrabian;R. Amin;P. Prucnal;T. El;V. Sorger
神经网络既需要输入的加权，也需要对输入和进行运算的非线性激活函数。神经网络加权已被证明在集成光子与干涉和基于环的波分复用。虽然在没有高光功率的情况下难以实现光学中的直接非线性，但是可以通过将光电二极管直接耦合到电光调制器来产生电光非线性。直接耦合组件的低电容导致工作速度>10 GHz，功耗相对较低。在这里，我们提出了一个封闭的形式方程的激活功能创建的石墨烯和量子阱电光吸收调制器电容耦合到光电二极管。我们基于调制器几何形状和热噪声的分析表明，这种电光神经元产生约60的SNR。在具有这些电光节点的前馈神经网络上执行MNIST分类推断测试，对于QW和基于石墨烯的调制器，分别以大约95%的精度开始大约5mW和20mW的激光功率水平。我们的研究结果显示，未来的光学和光子神经网络使用电光模拟（非尖峰）神经元的现实操作性能的区域。
Neural networks require both a weighting of inputs and a nonlinear activation function operating on their sum. Neural network weighting has been demonstrated in integrated photonics with both interferometric and ring-based wavelength division multiplexing. While direct nonlinearity in optics is difficult to achieve without high optical powers, an electro-optic nonlinearity can be created by directly coupling a photodiode to electro-optic modulator. The low capacitance of directly coupling the components results in operating speeds >10 GHz with relatively low power consumption. Here we present a closed form equation for the activation functions created by graphene and quantum well electro-optic absorption modulators capacitively coupled to photodiodes. Our modulator-geometry based and thermal-noise analysis shows that such electro-optic neurons produce SNRs around 60. Performing an MNIST classification inference test on a feed-forward neural network with these electrooptic nodes, with accuracies of about 95% starting a laser power level around 5mW and 20mW for the QW and Graphene-based modulator, respectively. Our findings show regions of realistic operating performance of future optical and photonic neural networks using electro-optic analogue (non-spiking)neurons.
影响因子：
35
作者：
Shen, Yichen;Harris, Nicholas C.;Soljacic, Marin
通讯作者：
Soljacic, Marin