白冰，裴丽，左晓燕
（北京交通大学，北京 100044）
摘要：提出了利用硅基光电子芯片进行人工神经网络计算处理的方法。硅基光电子芯片凭借光子的独特性质，能够在人工神经网络的计算处理中发挥高带宽、低时延等优势。在处理深度学习中大量的矩阵计算的乘加任务时，硅基光电子芯片拥有更高的处理速度和更低的能耗，从而有利于深度学习中的人工神经网络计算速度和性能的提升。
关键词：人工神经网络；硅基光电子芯片；人工智能；深度学习
Silicon Photonic Chips for Artificial Intelligence
BAI Bing, PEI Li, ZUO Xiaoyan
（Beijing Jiaotong University, Beijing 100044, China）
Abstract: Silicon photonic chips are used to perform artificial neural network computation. Because of the unique properties of photons, silicon photonic chips have the advantages of high bandwidth and low delay in the computation and processing of artificial neural network. When dealing with the multiplication and addition task of a large number of matrix calculations in deep learning, silicon photonic chips have higher processing speed and lower energy consumption, which is beneficial to the improvement of the computational speed and performance of artificial neural network in deep learning.
Keywords: artificial neural network; silicon photonic chips; artificial intelligence; deep learning
.png)
.png)
.png)
