This is a notebook I presented to a Deep Learning class I TA at Columbia University in October 2018. I hope to work a bit more on it and upload it as a good tutorial on TRFL since there are only a few resources on TRFL yet. The goal is to show how to assemble building blocks to solve OpenAI Gym games using RL, in particular using TRFL.

The current version is partially documented and available on nbviewer here.

One thing I'd like to do is to show how we can leverage available architecutres from keras-rl for example to use that as a building block for the network itself, therefore showing that with open source architectures + TRFL we can now write RL very easily.

Feel free to contribute!