Repository of raw data files and Matlab scripts required to reproduce results presented in “Gaussian Processes based Optimized Launch Power for Nonlinear Optical Fiber Transmission Links”

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Each dataset contains the received constellation signal to noise ratio as a function of the launch optical power into each span. Experiment: 10 spans, 100km of SSMF with launch VOA and post span EDFA. Signal 31.7 GBd PM-QPSK private signal modulated by Ciena WaveLogic3 and received by a 25 GHz ICR , 100GSa DSO, and private DSP. The data file in Matlab format 7.3, contains a single structure with: Results.Description - some notes Results.LP - launch powers into each span, and receiver [dBm] Results.mleSNR - constellation SNR [dB] Results.base_LP - prior knowledge or start point or sweep centre of optimum [dBm]

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

SingleChannelRandom_170919_R1_Public.mat Analyze with either GP_optimise_hypopt.m, GQ_optimise.m or PT_optimise.m There are three Matlab scripts to find the optimum launch power based on: GP - a Gaussian Process, GQ - a generalised Quadratic and PT - a physical transmission model. They have options for number of samples, GP_optimise_hypopt has options to optimise hyper-parameters and to include hyper-prior training. Used to produce figures 2,3,5 and 6

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

SingleChannelScan_200820_R4_Public.mat Analyze with PowerSweepCalc.m Used to produce figures 4 SingleChannelScan_290720_R2_Public.mat used to check PT result

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

SingleChannelBayesian_140720_R1_Public.m SingleChannelBayesian_140720_R2_Public.m SingleChannelBayesian_140720_R3_Public.m SingleChannelBayesian_140720_R5_Public.m SingleChannelBayesian_140720_R6_Public.m 5 repeat runs Analyze with PlotBayesianResults.m Used to produce figures 7 and 8

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

EmulatedSingleChannelBayesian_140720_Public.m BayesOpt.m MeasureSNR.m

provide an emulation of the Bayesian aquisition based optimisation. The experimental measurement of SNR is replaced by an emulation based on the PT model and provided by the MeasureSNR.m function.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

This work was supported by the UK Engineering and Physical Sciences Research Council through programme grant TRANSNET EP/R035342/1 and Ciena.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%