codingdaidai/LDBP

Physics-Based Deep Learning for Fiber-Optic Communication Systems

LDBP: Learned Digital Backpropagation

Getting Started

The code is based on TensorFlow 1.13.1 and may not work properly with other (older or newer) versions. It is recommended to create a dedicated conda environment using the YAML file in the folder conda as follows:

(base)~$ conda env create -f ldbp_env.yml
(base)~$ conda activate ldbp_env

Afterwards, it should be possible to run the provided jobscripts in the folder ldbp. For example:

(ldbp_env)~$ ./jobscript_isit

To train for different scenarios, most of the parameters and training options are set in a configuration file located in the folder config.

Additional Information

This repository is based on joint work with Henry D. Pfister. If you decide to use the source code for your research, please make sure to cite our paper(s):

• C. Häger and H. D. Pfister, "Physics-Based Deep Learning for Fiber-Optic Communication Systems", in IEEE J. Sel. Areas Commun. (to appear), 2020

• C. Häger and H. D. Pfister, "Nonlinear Interference Mitigation via Deep Neural Networks", in Proc. Optical Fiber Communication Conf. (OFC), San Diego, CA, March 2018

• C. Häger and H. D. Pfister, "Deep Learning of the Nonlinear Schrödinger Equation in Fiber-Optic Communication", In Proc. IEEE Int. Symp. on Information Theory (ISIT), Vail, CO, June 2018