光子脉冲神经元权重器件的研制

发布时间：2018-04-10 17:28

本文选题：光学器件 + 可调光衰减器　；参考：《中国激光》2017年04期

【摘要】：互连的光子脉冲神经元通过权重器件联系在一起,为了实现神经网络的大规模计算,权重器件的实现至关重要。利用微机电系统可调光衰减器(VOA),研制了一种可以自动调节光子脉冲神经元的权重器件。该权重器件包括VOA、光电探测器、单片机、模数转换器、数模转换器和放大器等模块,可以根据接收的光信号快速计算查表,可对VOA的衰减值进行实时在线调整。该权重器件效率高,且容易实现。该权重器件配合脉冲时间依赖的可塑性(STDP)光路使用,可以实现光子脉冲神经元的STDP学习机制。当STDP曲线窗口高度为0.2时,对权重器件进行了测量,实现了4种STDP学习。实验测量结果与理论计算结果一致。
[Abstract]:Interconnected photonic pulse neurons are connected by weight devices. In order to realize the large-scale calculation of neural networks, the realization of weight devices is very important.Using the adjustable optical attenuator (VOAA) of MEMS, a weight device which can automatically adjust the photon pulse neuron is developed.The weight device includes VOA, photodetector, MCU, A / D converter, digital-to-analog converter, amplifier and so on. It can quickly calculate and search the table according to the received optical signal, and can adjust the attenuation value of VOA in real time and online.The weight device is efficient and easy to realize.The STDP learning mechanism of the photonic pulsed neuron can be realized by using the weighted device in combination with the pulse time dependent plasticity (STDP) optical path.When the height of STDP curve window is 0.2, the weight devices are measured and four kinds of STDP learning are realized.The experimental results are in agreement with the theoretical results.
【作者单位】：北京交通大学理学院光信息科学与技术研究所发光与光信息技术教育部重点实验室;北京交通大学电气工程学院;
【基金】：国家自然科学基金(61571035,61378061,61401017)
【分类号】：TN15

【相似文献】