Does RL work in practice? I have seen many impressive applications of end-to-end deep learning, but nothing impressive at all when it comes to RL in the real world (especially when compared to control theory in robotics domain).
Great book from one of my favourite profs at UofA.
It's very focused and provides examples on how to implement what I would call classic RL algorithms, and can serve as a decent intro. Don't expect to read it and implement Deep RL right away however.