TL;DR;

This is the third course I have taken in the OMSCS program (Fall 2022). In general, I love it so much. It consists of six programming exercises with three bigger progjects. It covers both traditional reinforcement learning techniques and deep reinforcement learning frontier methods. Most of the these exercises are engaging and fun to play with.

Contents

There are many people discussed the contents of already, so, I would like to keep this part short.

  1. Focus on the writting. All projects involves a lot of programming while the grading is merely based on your report. You may want to follow instructions closely.
  2. Read the paper first before dive into the coding part. You would use Python, which would not require you write tons of code. However, you should know what you are doing.
  3. The last project is really time consuming (my version is a football playing game). You would implement an agent to play with the baseline agent. The whole game is to play around with the enviroment, implement plausible algorithms and generate enough plots. As you may know, train a deep RL agent takes a lot of time. So, be prepare for that.

Other than that, assignements are relatively short and easy, which are mainly used to help you understand the key idea behind the scene.

Last advice, the lectures’ quality is relatilvey low. So, I would recommend you to read the textbook closely. And watch some YouTube videos to strengthen your understanding about all the topics.

But in general, you would learn a lot by doing, which is core of computer science.

My Thoughts

If you want to learn more about another branch of AI, this is a highly recommended course. Most of the materials are useful in work and daily uses. The first half of the course would pave your road towards understanding the basics and the later half would help you to keep up the speed of current RL development.