Meta-learning Accelerated Bi-LSTM for Fiber Nonlinearity Compensation

Fiber nonlinearity is one of the main limitations for long-reach optical fiber transmissions. We propose a meta-learning accelerated Bi-LSTM algorithm for fiber nonlinearity equalization by saving the training time with different launch powers. The proposed algorithm obtains 0.7 dB Q-factor gain com...

Full description

Saved in:
Bibliographic Details
Published in2023 21st International Conference on Optical Communications and Networks (ICOCN) pp. 1 - 3
Main Authors Ren, Xuecheng, Liu, Jiaming, Huang, Xiatao, Zhang, Qianwu, Zhang, Jing, Qiu, Kun
Format Conference Proceeding
LanguageEnglish
Published IEEE 31.07.2023
Subjects
Bi-LSTM
Complexity theory
Equalizers
Meta-learning
Metalearning
Nonlinear equalization
Optical fiber communication
Optical fiber networks
Optical fibers
Q-factor
Training
Online AccessGet full text

Cover

More Information
Summary:Fiber nonlinearity is one of the main limitations for long-reach optical fiber transmissions. We propose a meta-learning accelerated Bi-LSTM algorithm for fiber nonlinearity equalization by saving the training time with different launch powers. The proposed algorithm obtains 0.7 dB Q-factor gain compared to DBP and 81.25% complexity reduction.
ISSN:2771-3059
DOI:10.1109/ICOCN59242.2023.10236105