A low complexity nonlinear compensation method based on multiple-input kubota operators

By employing a low-complexity nonlinear compensation method based on multi-input Koupman operators and utilizing the MiKNO neural network to approximate the solution space of the DBP algorithm, the problem of high computational complexity in optical fiber communication is solved, achieving efficient nonlinear compensation.

CN116667931BActive Publication Date: 2026-06-09SOUTHWEST JIAOTONG UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
SOUTHWEST JIAOTONG UNIV
Filing Date
2023-05-31
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing digital backpropagation (DBP) and enhanced DBP (LDBP) algorithms suffer from high computational complexity in optical fiber communication, especially in transoceanic transmission where efficient nonlinear compensation is difficult to achieve.

Method used

A low-complexity nonlinear compensation method based on multi-input Koopman operators is adopted. By constructing a multi-input Koopman operator (MiKNO) neural network, the solution space of the DBP algorithm is approximated by the Koopman operator, thereby reducing computational complexity and achieving nonlinear compensation.

Benefits of technology

It significantly reduces computational complexity in transoceanic transmission, achieving compensation performance similar to DBP and LDBP, while adapting to different link parameter variations. The computation time decreases exponentially, making it suitable for nonlinear compensation over 12,000 km.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116667931B_ABST
    Figure CN116667931B_ABST
Patent Text Reader

Abstract

The application discloses a low-complexity nonlinear compensation method based on a multi-input Kupman operator, and specifically comprises the following steps: firstly, transmission experiments are carried out based on a fiber transmission experiment platform, and training sets are constructed by collecting experimental data from a real-time oscilloscope; then, a neural operator is constructed by using a neural network; finally, the weight parameters of the multi-input Kupman neural operator are iteratively updated by using training data to complete approximation of a digital back propagation algorithm. Compared with the digital back propagation algorithm, the application saves a large number of iteration steps, greatly improves the calculation efficiency, realizes 12000km nonlinear compensation, can cope with the change of nonlinear effects caused by different link parameters, and does not need to repeatedly train for different transmission scenes.
Need to check novelty before this filing date? Find Prior Art