/B2DQN

Code for my publication: Efficient Exploration through Bootstrapped and Bayesian Deep Q-Networks for Joint Power Control and Beamforming in mmWave Networks. Paper accepted for publication to IEEE Communications Letters.

Primary LanguageJupyter Notebook

Bootstrapped and Bayesian Deep Q-Networks

How to use

  • Set the number of antennas in the base station. In environment.py change the line self.M_ULA to the values of your choice. The code expects M = 4, 8, 16, 32, and 64.
  • Run DQN variants algorithms. Run the scripts DQN, BoDQN, BaDQN, and B2DQN.py in folder Codes. The result is the same as that in folder Results.
  • Show the results. Run the script Results_plot.ipynb in folder Results to show Figure 3, Figure 4, and Table IV in the paper.