基于深度学习的常规单模光纤信道模拟
首发时间:2020-01-14
摘要:为了简化验证光纤链路中数据传输的操作,以通信系统仿真软件(VPI)作为数据采集来源,在实际80km G.652单模光纤结果数据下,控制其眼图、波形图等数据逼近真实值,采得数据,后在TensorFlow平台,采用深度学习的方法,其中主要采用RNN(递归神经网络)中LSTM(长短时记忆)的方法,并以DNN(深度神经网络)作为实验对比方法,调整参数得出千组实验结果,取其中4组作为最后结果,得出LSTM的适应性良好,DNN适应性较差,对深度学习在光纤信号数据模拟预测中这一想法验证成功。
For information in English, please click here
Optical Fiber Signal Simulation Based on Deep Learning
Abstract:In order to simplify the operation of verifying the data transmission in the optical fiber , the communication system simulation software (VPI) is used as the data collection source. Under the actual 80km G.652 single-mode fiber result data , the eye diagram, waveform diagram and other data are controlled to approximate the real value and acquire the data.Then use the deep learning method on the TensorFlow platform, which mainly uses the LSTM (Long and Short Time Memory) method in RNN (Recurrent Neural Network), and use DNN (Deep Neural Network) as the experimental comparison method. The parameters were adjusted to get thousands of experimental results, and four of them were taken as the final results. It was concluded that the adaptability of LSTM is good and the adaptability of DNN is poor. The idea of deep learning in the simulation and prediction of fiber signal data was successfully verified.
Keywords: Communication and information system Deep learning Optical fiber simulation
引用
No.****
动态公开评议
共计0人参与
勘误表
基于深度学习的常规单模光纤信道模拟
评论
全部评论0/1000