Yesterday morning, for all the reasons I outlined in the last post, I was torn about whether to sign up for the Udacity Deep Reinforcement Learning Nanodegree. So I went to the main course page for enrollment, and for some reason, what used to be a $1436 course can now be had for $699.

I know that this may well be a deliberate trick. Who cares. I signed up immediately and am very glad that I did. So far, I’m loving the detail with which they’re reviewing the foundational elements of RL. I had already had a basic understanding of things like Markhov Decision Processes and discounted reward functions. But it’s very useful to see a step-by-step walkthrough of exactly how they work backed up by notes, videos, and quizzes.

My favorite part of the experience so far is the mathematical notation. On the one hand, I’ve never had a problem understanding the concepts behind various AI principles. I’ve relied on that pretty heavily in educating myself on what I need to learn. However, to read research papers and really understand them it helps to know how to read the notation. So I’m quite pleased that a significant portion of the course (at least early on) seems to be devoted to taking the concepts they teach and explicitly mapping each part of a concept to any mathematical notation that expresses those concepts. That alone will be hugely valuable well beyond what gets covered in the coursework.