摘要
针对梯度下降(GD)算法收敛速度较慢的问题,提出了一种基于GD优化算法的有限脉冲响应(FIR)色散均衡器。采用动量梯度下降(MGD)算法在每次更新迭代时对梯度进行指数加权平均优化(简称CMGD算法)并进行偏差修正,利用Opti System搭建单载波相干光传输系统并与Matlab联合仿真。仿真结果表明:在学习速率为0.1、动量系数为0.98时,CMGD算法收敛值达到0.015461;在不同信噪比条件下,采用CMGD算法设计的均衡器的均衡效果优于GD-FIR和经典时域FIR的均衡器,且降低了硬件实现复杂度。
Aiming at the slow convergence speed of gradient descent(GD) algorithm,a finite impulse response(FIR) dispersion equalizer based on GD optimization algorithm is proposed.The momentum gradient descent(MGD) algorithm is used to optimize the gradient exponentially weighted average(CMGD algorithm for short) and correct the deviation at each update iteration.The single carrier coherent optical transmission system is built by optisystem and simulated with MATLAB.The simulation results show that when the learning rate is 0.1 and the momentum coefficient is 0.98,the convergence value of CMGD algorithm reaches 0.015461.Under different SNR conditions,the equalization effect of the equalizer designed by CMGD algorithm is better than that of GD-FIR and classical time-domain FIR,and the hardware implementation complexity is reduced.
作者
钱正成
陈睿
陆叶
李传起
周省邦
谢乙田
孔一卜
QIAN Zhengcheng;CHEN Rui;LU Ye;LI Chuanqi;ZHOU Shengbang;XIE Yitian;KONG Yibu(College of Electronic Engineering,Guangxi Normal University,Guilin Guangxi 541004,China;Colloge of Physics and Electronic Engineering,Nanning Normal University,Nanning 530001,China;Guangxi Radio and TelevisionInformation Network Co.,Ltd.,Nanning 530023,China;The 34th Research Institute of CETC,Guilin 541004,China)
出处
《光通信技术》
2022年第2期85-90,共6页
Optical Communication Technology
基金
广西科技计划项目(AB17292082)资助。
关键词
相干光通信
色散均衡
动量梯度下降
有限脉冲响应均衡器
抽头权值
coherent optical communication
dispersion equalization
momentum gradient descent
finite impulse response equalizer
tap weight
