This directory contains the source-code used for SI cancellation using model-based NNs and complex-valued neural networks. The directory contains the following sub-directories:

• applications: Contains the source code for specific applications, e.g. full-duplex cancellation.
• complex_nn: Main source code for the project, e.g., custom definitions of layers, initializers, and so forth.

The main approach for the complex-valued neural networks is as described here here and here.

To setup a conda environment just run (in this directory)

conda env create -f env_rf_unfolding.yml

This will install all the required packages, including TensorFlow v2.

Alternatively, the following set of commands will provide the necessary packages using Anaconda. This assumes that conda-forge has been added, otherwise it can be added using the command

conda config --append channels conda-forge

To setup the environment, call (you may also start by cloning base conda create --name env_rf_unfolding --clone base)

conda create -n env_rf_unfolding python=3.7
conda activate env_rf_unfolding
conda install -c anaconda pandas
conda install scipy
conda install -c conda-forge matplotlib
conda install seaborn scikit-learn
conda install -c conda-forge tqdm
conda install -c anaconda tensorflow