kimdoohan's honest review, From Introduction to Reinforcement Learning to Deep Q-learning/Policy Gradient course
Reviews 2
Average rating 5
The only Korean-language course with a “good balance of theory and practice” <Pros> 1. It doesn’t get bogged down in deriving equations from the very beginning. Instead, it explains the concepts using a reasonable amount of math and then moves straight to implementation, which makes it great for learning by doing. 2. It pairs extremely well with Hyeokpenheim’s reinforcement learning theory lectures on YouTube (some of which are paid). 3. Beginner-friendly reinforcement learning courses are very rare. They tend to offer either theory alone or code with no explanation. In that regard, this course, with its balanced mix of theory and implementation, was personally the most helpful. I took three courses—the KOCW open course, Hyeokpenheim’s theory course, and this one—and this was the best. <Cons> The policy-based learning section only goes as far as REINFORCE. It would be perfect if it covered Actor-Critic, A2C, DDPG, and PPO as well. However, if you understand the generalized policy gradient equation and know REINFORCE as an example, the other algorithms are only minor? variations. So I think it should be entirely possible to find resources on your own and implement them using what you learned in the course. Thank you. I learned a lot. This was the only proper Korean-language reinforcement learning course that included both theory and practice. I felt overwhelmed when I only watched other theory courses, but after taking this one, I think I can now study and implement individual algorithm papers on my own.

