Abstract
A method is proposed to select the optimal power(s) of training data for learned digital backpropagation (LDBP): Low-step LDBP should be trained on multi-power data, while high-step LDBP should be trained on high single-power data.
© 2021 The Author(s)
PDF ArticleMore Like This
Zhen Wu, Du Tang, Zhongliang Sun, and Yaojun Qiao
T4A.71 Asia Communications and Photonics Conference (ACP) 2021
Du Tang, Zhen Wu, Xizi Tang, Jiating Luo, Ji Luo, Bofang Zheng, and Yaojun Qiao
W4I.1 Optical Fiber Communication Conference (OFC) 2022
Xiangyu Yang, Du Tang, Zhen Wu, and Yaojun Qiao
T4A.114 Asia Communications and Photonics Conference (ACP) 2021