avinashavuthu/deep-reinforcement-learning-for-5g-networks

Code for my publication: Deep Reinforcement Learning for 5G Networks: Joint Beamforming, Power Control, and Interference Coordination. Paper accepted for publication to IEEE Transactions on Communications.

Deep Reinforcement Learning for 5G Networks

How to use

The code to run voice is self explanatory.

For data, start by creating a folder figures in the same directory as your fork. 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.

For optimal, uncomment lines 428 and 437 from main.py. Comment out lines 426, 439, 440, 442. When run is complete, rename the figures folder to become figures M=m optimal after completion, where m takes values of M as shown above.

For the proposed solution, uncomment lines 426, 439, 440, 442 from main.py. Comment out lines 428 and 437. When run is complete, rename the figures folder to become figures M=m.

Run the script parse.py in every folder figures* you create. This generates a few intermediary files.

Create a folder figures again. Now run plotting.py. If you have any problems related to LaTeX plotting, change all the lines matplotlib.rcParams['text.usetex'] = True to matplotlib.rcParams['text.usetex'] = False then re-run.

For reproducibility, please use CPU and not the GPU when running the code.

Version history

6/28/2019 Initial code release

11/6/2019 Version 2. Normalized the power and the convergence episodes. I choose the episode close to the median to determine convergence.

12/15/2019 Version 2.1. Introduced the optimal solution for voice.