Machine Learning with Phil has got another interesting look at Deep Q Learning as part of a preview of his course.
The two biggest innovations in deep Q learning were the introduction of the target network and the replay memory. One would think that simply bolting a deep neural network to the Q learning algorithm would be enough for a robust deep Q learning agent, but that isn’t the case. In this video I’ll show you how this naive implementation of the deep q learning agent fails, and spectacularly at that.
This is an excerpt from my new course, Deep Q Learning From Paper to Code which you can get on sale with this link