Research Data Supporting "Fast Neural Network Inverse Model to Maximize Throughput in Ultra-Wideband WDM Systems"

This dataset (one excel file with multiple sheets) contains all the necessary data used to produce Figure 1 and Figure 4 to 10 in our manuscript. It contains the fibre attenuation and Raman gain spectrum used in the physical layer for Figure 1; hyperparameter optimisation, training, and validation data for the proposed DNN for Figure 4; performance evaluation of 3 approaches mentioned in the manuscript for Figure 5; optimal launch power profiles by PSO and the corresponding SNR profiles for Figure 6; Flat target SNR profiles and the desired launch power profiles for Figure 7; Parameter optimisation for the proposed greedy algorithm for Figure 8; optimal launch power profiles by the greedy algorithm and the corresponding SNR profiles for Figure 9; and achieved throughput and computational time of all approaches for Figure 10.

本数据集为单份含多个工作表的Excel文件，包含了本文稿件中图1及图4至图10制作所需的全部必要数据。其中涵盖：用于图1物理层的光纤衰减与拉曼增益光谱数据；本文提出的深度神经网络（Deep Neural Network, DNN）用于图4的超参数优化、训练与验证数据集；本文提及的3种方法的性能评估数据（对应图5）；粒子群优化（Particle Swarm Optimization, PSO）得到的最优发射功率曲线及其对应信噪比（Signal-to-Noise Ratio, SNR）曲线数据（对应图6）；平坦化目标信噪比曲线与期望发射功率曲线数据（对应图7）；本文提出的贪心算法用于图8的参数优化数据；贪心算法得到的最优发射功率曲线及其对应信噪比曲线数据（对应图9）；以及所有方法的实现吞吐量与计算耗时数据（对应图10）。