Optimal Solutions for Joint Beamforming and Antenna Selection: From Branch and Bound to Machine Learning
Implementation of the paper "Optimal Solutions for Joint Beamforming and Antenna Selection: From Branch and Bound to Graph Neural Imitation Learning", published in IEEE Transactions on signal processing. Also, see a shorter version in this ICASSP Paper.
Make sure that you have pip installed.
If you are on linux. Run the following command
./initial_setup.bash
Else, manually create the following directories
antenna_selection/data_bf/data
antenna_selection/data_rbf/data
antenna_selection/data_bf/trained_models
antenna_selection/data_rbf/trained_models
And run
pip install -r requirements.txt
On the base directory run the following to make sure that the repo is added to PATH and PYTHONPATH environment variables
source activate_env.bash
Finally, you can run the proposed B&B procedure by running the following
python antenna_selection/bb_unified.py
You can train node classifier with the following
python models/dagger_multiprocess.py
Change the parameters in models/setting.py and on the individual file that you are running.