This repo is the implementation of the paper Improving the Diversity of Bootstrapped DQN via Noisy Priors on ALE games.
atari-py installed from https://github.com/kastnerkyle/atari-py
torch='1.0.1.post2'
cv2='4.0.0'
matplotlib='3.3.2'
run_bootstrap.py
is the main file. visual_plot
is the file to plot results.
To run experiment, set info
in run_bootstrap.py
. To run different games, set game names in info['GAME']
. Set info['IMPROVEMENT']
to use noisy priors or not.
python run_bootstrap.py
Credit to which the project is built on: