In situ optical backpropagation training of diffractive optical neural networks  

作　　者：TIANKUANG ZHOU LU FANG TAO YAN JIAMIN WU YIPENG LI JINGTAO FAN HUAQIANG WU XING LIN QIONGHAI DA 

机构地区：[1]Department of Automation,Tsinghua University,Beijing 100084,China [2]Institute for Brain and Cognitive Science,Tsinghua University,Bejing 100084,China [3]Tsinghua Shenzhen International Graduate School,Tsinghua University,Shenzhen 518055,China [4]Bejing Innovation Center for Future Chip,Tsinghua University,Bejing 100084,China [5]Institute of Microelectronics,Tsinghua University,Bejjing 100084,China [6]Bejing National Research Center for Information Science and Technology,Tsinghua University,Bejing 100084,China

出　　处：《Photonics Research》2020年第6期940-953,共14页光子学研究（英文版）

基　　金：Beijing Municipal Science and Technology Commission(No.Z181100003118014);National Natural Science Foundation of China(No.61722209);Tsinghua University Initiative Scientific Research Program.

摘　　要：Training an artificial neural network with backpropagation algorithms to perform advanced machine learning tasks requires an extensive computational process.This paper proposes to implement the backpropagation algorithm optically for in situ training of both linear and nonlinear diffractive optical neural networlks,which enables the acceleration of training speed and improvement in energy efficiency on core computing modules.We demonstrate that the gradient of a loss function with respect to the weights of diffractive layers can be accurately calculated by measuring the forward and backward propagated optical fields based on light reciprocity and phase conjunction principles.The diffractive modulation weights are updated by programming a high-speed spatial light modulator to minimize the error between prediction and target output and perform inference tasks at the speed of light.We numerically validate the effectiveness of our approach on simulated networks for various applications.The proposed in situ optical learning architecture achieves accuracy comparable to in silico training with an electronic computer on the tasks of object dlassification and matrix-vector multiplication,which further allows the diffractive optical neural network to adapt to system imperfections.Also,the self-adaptive property of our approach facilitates the novel application of the network for all-optical imaging through scattering media.The proposed approach paves the way for robust implementation of large-scale difractive neural networks to perform distinctive tasks all-optically.

关 键 词：process. WEIGHTS BACKWARD 

分 类 号：O43[机械工程—光学工程] TP183[理学—光学]