This is the Code for "Deep Q Learning - The Math of Intelligence #9" By Siraj Raval on Youtube
This weeks challenge is to use Q learning to train an agent for any game you'd like. You can use OpenAI's Gym or Universe as a simulation testbed, but for the Q-Learning algorithm itself don't use any libraries. Bonus points if you use a deep convolutional network from scratch as well to learn from pixels, which means your agent is generalized to more than just one game (Deep Q Learning). Good luck!
This is the code for this video on Youtube by Siraj Raval as part of the Math of Intelligence series. We're going to rerecreate DeepMind's Deep Q Learner for a variety of games.
- keras (http://www.pyimagesearch.com/2016/11/14/installing-keras-with-tensorflow-backend/)
- tensorflow (https://www.tensorflow.org/install/)
- gym (https://github.com/openai/gym)
- collections
Use pip to install any dependencies.
Just run jupyter notebook in terminal and the code will run. If you'd like to run this code on Super Mario, you need to install this additonal dependency.
The credit for this code goes to PeterWittek. I've merely created a wrapper to get people started.