Introduction

This package provides a Theano-based implementation of the deep Q-learning algorithm described in:

Playing Atari with Deep Reinforcement Learning Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller

The neural network code is largely borrowed from Sander Dieleman's solution for the Galaxy Zoo Kaggle challenge.

The results obtained with this code are not quite as good as the results from the paper. This could just be a matter of parameter tuning (I've done very little) or it could be something more fundamental.

Here is a video showing a trained network playing breakout:

http://youtu.be/SZ88F82KLX4

Dependencies

• A reasonably modern NVIDIA GPU
• Cython
• OpenCV
• Theano
• Pylearn2
• Arcade Learning Environment This package requires a slightly modified version of ALE. You'll need to replace rlglue_controller.cpp with the provided version before compiling. This version handles down-sampling the image and converting to gray-scale.
• RL-Glue
• RL-Glue Python Codec

Running

Use the script ale_run.py to start all the necessary processes:

$ python ale_run.py --exp_pref data

This will store output files in a folder prefixed with data in the current directory. Pickled version of the network objects are stored after every epoch. The file results.csv will contain the testing output. You can plot the progress by executing plot_results.py:

$ python plot_results.py data_09-29-15-46_0p0001_0p9/results.csv

After a couple of days, you can watch the trained network play using the ale_run_watch.py script:

$ python ale_run_watch.py data_09-29-15-46_0p0001_0p9/network_file_99.pkl

See Also

• https://github.com/kristjankorjus/Replicating-DeepMind

  This is a work in progress. Their project is described here: http://robohub.org/artificial-general-intelligence-that-plays-atari-video-games-how-did-deepmind-do-it/

• https://github.com/brian473/neural_rl

  This is an almost-working implementation developed last spring by my student Brian Brown. I haven't reused his code, but Brian and I worked together to puzzle through some of the blank areas of the original paper.