dortsur/capacity-estimator-via-dine

Capacity Estimation via Directed Information Neural Estimator (DINE)

This repository contains an implementation of capacity estimator of continuous channels as introduced in (Link will be added in the future).

Prerequisites

The code is compatible with a tensoreflow 2.0 environment. If you use a docker, you can pull the following docker image

docker pull tensorflow/tensorflow:latest-gpu-py3

Running the code

The estimate the capacity of the AWGN channel run

python ./main.py --name <simulation_name> --config_name awgn --P <source_power> --C <capacity_for_visualization> &

The estimate the capacity of the MA(1)-AGN channel run

python ./main.py --name <simulation_name> --config_name arma_ff --P <source_power> --C <capacity_for_visualization> &

The estimate the feedback capacity of the MA(1)-AGN channel run

python ./main.py --name <simulation_name> --config_name arma_fb --P <source_power> --C <capacity_for_visualization> &

Authors

• Ziv Aharoni
• Dor Tsur
• Ziv Goldfeld
• Haim Permuter

License

This project is licensed under the MIT License - see the LICENSE.md file for details