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.

Beamforming and Antenna Selection Problem

Proposed Solution

Results

Setup

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

Executing the Code

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.