This repository contains an implementation of Deep Q Network that plays the Atari game "Breakout".
It is much inspired by this implementation

https://github.com/matthiasplappert/keras-rl

The project has been written in Keras and Python and uses OpenAI Gym and ALE.

The links below is the ones that I have gotten the most from.

https://www.intelnervana.com/demystifying-deep-reinforcement-learning/

http://robohub.org/artificial-general-intelligence-that-plays-atari-video-games-how-did-deepmind-do-it/

https://keon.io/deep-q-learning/

https://www.theguardian.com/global/2017/mar/14/googles-deepmind-makes-ai-program-that-can-learn-like-a-human

https://jaromiru.com/2017/05/27/on-using-huber-loss-in-deep-q-learning/

https://github.com/matthiasplappert/keras-rl

https://github.com/openai/gym

https://openai.com/research/

https://techcrunch.com/2014/01/26/google-deepmind/

https://deepmind.com/research/publications/playing-atari-deep-reinforcement-learning/

https://storage.googleapis.com/deepmind-media/dqn/DQNNaturePaper.pdf

To just run the code, clone it and cd into dqn folder. Then write

"python breakout.py --mode run"

Now the game will play with the trained model.

To train the model, write "python breakout.py --mode train".

For installing libraries, check out the installation instructions here:

https://github.com/matthiasplappert/keras-rl