Research data supporting "Physics-based Modeling for Hybrid Data Driven Models to Estimate SNR in WDM Systems"

The Excel file contains 7 sheets in total; they are associated to figures 2 to 7, and 9, respectively. The processed data was obtained by analysing 700 datapoints of 15 launch powers each, in addition to 700 datapoints of the corresponding resulting 15 SNR values after propagation in the experimental setup described in the paper. Sheet for Fig 2 contains the number of training datapoints and their corresponding overall root mean square error (RMSE) for each of the methods described in the paper. Sheet for Fig. 3. includes the SNR estimation error and their frequency of occurrence for each of the methods. Sheet for Fig 4 contains the average XPM and SPM values based on channel distance for each of the methods, and the values used for the plotting of the fit based on method 2. Sheet for Fig 5 contains the value of the per channel RMSE per method. Sheet for Fig 6 includes the per channel back to back SNR values computed from dividing the data into groups of 100 and applying each of the methods as described in the paper. Sheet for fig 7 contains the number of training datapoints and the corresponding RMSE for each of physical model 2, neural network (NN), and Gaussian process regression (GPR). Sheet for Fig 9 contains the number of training datapoints and the corresponding RMSE values for the GPR, and for the hybrid model when using 100 or 10 for the physical model part.