GP Driven DBP

Change log
Nevin, Josh 
Savory, Seb 
Polina, Bayvel 
Galdino, Lidia 
Sillekens, Eric 
  1. containing waveforms generated from experiments : This dataset contains 30 sample waveforms generated from the experimental system of a wavelength division multiplexing (WDM) optical network link. This link is with 4 channels, of 1000 km length, of 64GBd symbol rate and 64 QAM modulation format. They are in the format of ".mat" and can be viewed with Matlab. $$ \ $$2) containing simulation results from the proposed Gaussian process (GP) driven digital backpropagation (DBP) approach : This dataset contains the resultant data generated from the simulations which are then used to generate the plots in the paper. Data files with respect to signal to noise ratio (SNR) data are represented in “.mat” format and can be accessed with Matlab while the other data files are in the format of “.pkl” and can be opened using Python language “pickle” module. The trained GP models are also stored in “.pkl” format and can be accessed as described above.$$ \ $$3) containing all source files and remaining data: These files contain the source codes related to this work including GP implementation, DBP simulation and history matching. Sources are written in Python programming language and can be executed with any Python environment. Additionally, this contains datasets generated from the optimisation process in “.mat” format. See the main manuscript for more details.

Software / Usage instructions
Digital Backpropogation, Gaussian Processes, Optical Networks
Engineering and Physical Sciences Research Council (EP/R035342/1)
EPSRC Transnet grant number EP/R035342/1