数据驱动型光纤中短距离传输预测模型研究数据

该数据来源于数据驱动型光纤模型的研究中，其具体为光纤传输预测模型结构中的多头注意力权值分布，以10GBaud OOK和40GBaud PAM信号的传输为例，数据由Python IDE生成并可由Matlab进行查看与进一步处理。通过将广泛应用于Transformer中的多头注意力机制引入光纤模型中，模型能够在提升输入输出波形特征提取精度的基础上进一步增加距离泛化的性能。模型在获得有效训练后，能够有效地预测100km以内任意距离下的信号传输。该模型中的预测模型发表于Optics Express上，接收模型发表于COL上。支撑论文“Data-driven fiber model based on the deep neural network with multi-head attention mechanism”、“Fiber communication receiver models based on the multi-head attention mechanism”以及专利“一种光纤传输信号预测模型的训练方法及装置”和“一种光通信接收模型的训练方法及装置”该模型优化后可应对“百公里”级传输需求。

This dataset is sourced from research on data-driven optical fiber models, specifically focusing on the multi-head attention weight distributions within the architecture of optical fiber transmission prediction models. Taking the transmission of 10GBaud OOK and 40GBaud PAM signals as examples, the dataset was generated using a Python IDE and can be visualized and further processed with MATLAB. By integrating the multi-head attention mechanism, which is widely adopted in Transformer models, into the optical fiber model, the proposed model can improve the accuracy of feature extraction for input and output waveforms while enhancing its distance generalization capability. After effective training, the model can accurately predict signal transmission at any distance within 100 km. The prediction model in this work was published in Optics Express, while the receiver model was published in COL. This work is supported by two academic papers: "Data-driven fiber model based on the deep neural network with multi-head attention mechanism" and "Fiber communication receiver models based on the multi-head attention mechanism", as well as two patents: "Training Method and Apparatus for Optical Fiber Transmission Signal Prediction Model" and "Training Method and Apparatus for Optical Communication Receiver Model". After optimization, the model can meet the transmission requirements for 100-kilometer-scale optical links.

搜集汇总

背景与挑战

背景概述

该数据集聚焦于光纤中短距离传输预测模型研究，包含多头注意力权值分布数据，以10GBaud OOK和40GBaud PAM信号为例，支持100km内信号传输的预测。数据由Python生成，可用Matlab处理，模型通过引入多头注意力机制提升了特征提取精度和距离泛化性能，并支撑了相关论文和专利，适用于'百公里'级光纤通信优化需求。