November 15, 2020

265 words 2 mins read



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.

repo name farismismar/Deep-Reinforcement-Learning-for-5G-Networks
repo link
language Python
size (curr.) 11980 kB
stars (curr.) 60
created 2019-03-10

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 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 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 Comment out lines 428 and 437. When run is complete, rename the figures folder to become figures M=m.

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

Create a folder figures again. Now run 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.

comments powered by Disqus