( Deep) RL Stream, 2: Lets (fail to) make a system that solves a maze by looking at it

Continuation of the last reinforcement learning stream, we re going to reformulate the maze class so that it can output images for every state. Then we will put together and train a Deep RL based agent that will only receive an image of the maze rather than being given its internal state. I will probably use Keras to design and fit the neural net. So excited to learn how to do this :). SOME OF MY OTHER VIDOES: Stream, 1: Explaining RL to a baby: www. y