This software is a python implementation of Deep Q-Networks for playing ATARI games with Chainer package.

I followed the implementation described in:

• V. Mnih et al., "Playing atari with deep reinforcement learning"

http://arxiv.org/pdf/1312.5602.pdf

• V. Mnih et al., "Human-level control through deep reinforcement learning"

http://www.nature.com/nature/journal/v518/n7540/abs/nature14236.html

For japanese instruction of DQN and historical review, please check:

http://qiita.com/Ugo-Nama/items/08c6a5f6a571335972d5

My implementation is dependent on RL-glue, Arcade Learning Environment, and Chainer. To run the software, you need following softwares/packages.

• Python 2.7+
• Numpy
• Scipy
• Chainer (1.3.0): https://github.com/pfnet/chainer
• RL-glue core: https://sites.google.com/a/rl-community.org/rl-glue/Home/rl-glue
• RL-glue Python codec: https://sites.google.com/a/rl-community.org/rl-glue/Home/Extensions/python-codec
• Arcade Learning Environment (version ALE 0.4.4): http://www.arcadelearningenvironment.org/

This software was tested on Ubuntu 14.04 LTS.

Please check readme.txt