
I wrote this post earlier this year but never came around to hitting the publish button. I hope it can be useful as an intro to self reinforcement learning and combining that with neural networks. I'm also using another dataset than the typical toy grid worlds, which hopefully is refreshing:-)