Efficient stochastic parallel gradient descent training for on-chip optical processor
芯片光处理器的高效随机并行梯度下降训练
オンチップ光プロセッサの効率的なランダム並列勾配降下訓練
편상광 프로세서의 고효율 무작위 병렬 경도 하강 훈련
Entrenamiento de descenso de gradiente paralelo aleatorio eficiente para procesadores ópticos en chip
Formation de descente de gradient parallèle aléatoire efficace pour les processeurs optiques sur puce
Эффективное рандомизированное параллельное снижение градиента процессора на пластине
Yuanjian Wan 万远剑 ¹ ², Xudong Liu 刘旭东 ¹ ², Guangze Wu 吴广泽 ¹ ², Min Yang 杨敏 ¹ ², Guofeng Yan 颜国锋 ¹ ², Yu Zhang 张宇 ¹ ², Jian Wang 王健 ¹ ²
¹ Wuhan National Laboratory for Optoelectronics and School of Optical and Electronic Information, Huazhong University of Science and Technology, Wuhan 430074, China
中国 武汉 华中科技大学光学与电子信息学院 武汉光电国家研究中心
² Optics Valley Laboratory, Wuhan 430074, China
中国 武汉 湖北光谷实验室
In recent years, space-division multiplexing (SDM) technology, which involves transmitting data information on multiple parallel channels for efficient capacity scaling, has been widely used in fiber and free-space optical communication systems. To enable flexible data management and cope with the mixing between different channels, the integrated reconfigurable optical processor is used for optical switching and mitigating the channel crosstalk.
However, efficient online training becomes intricate and challenging, particularly when dealing with a significant number of channels. Here we use the stochastic parallel gradient descent (SPGD) algorithm to configure the integrated optical processor, which has less computation than the traditional gradient descent (GD) algorithm. We design and fabricate a 6×6 on-chip optical processor on silicon platform to implement optical switching and descrambling assisted by the online training with the SPDG algorithm.
Moreover, we apply the on-chip processor configured by the SPGD algorithm to optical communications for optical switching and efficiently mitigating the channel crosstalk in SDM systems. In comparison with the traditional GD algorithm, it is found that the SPGD algorithm features better performance especially when the scale of matrix is large, which means it has the potential to optimize large-scale optical matrix computation acceleration chips.
