Reinforcement Learning

RL Courses

Prerequisites: Foundation courses.
Recommended path
    The most approachable introduction to RL is the UAlberta MOOC, which is taught in a really digestible manner.
    For a more mathematically inclined RL introduction, go either for Berkeley's CS285 or Stanford's CS234 course and use UCL/DeepMind's course as complementary lectures. To choose between CS285 or CS234, watch the first lecture of each course to figure out which lecture style you prefer and take into consideration whether you want a stronger focus on deep RL (CS285).
Course
Year
Description
Difficulty Level
Resources
2020
As the name of class indicates and Sergey Levine makes clear in the first lecture, this course is concerned with deep RL. While a lot of material intersects with CS234, it is generally more DL-oriented (e.g. the discussed examples). It also goes into topics such as recent algorithmic questions and open problems in deep RL.
Medium
2019
The course spends is less centered around deep RL than CS285 (although the line between deep/non-deep can be quite blurry). The lecture notes are very extensive and the course site also contains a midterm review. Both courses are very well-taught, so the choice between these two courses comes down to personal preference.
Medium
β€‹πŸ€–β€‹πŸ—’οΈβ€‹
UAlberta - RL specialization​
2019
This 4-part MOOC is the most approachable RL course online. Topics include Markov Decision Processes, Dynamic Programming, Temporal Difference Learning and Policy Gradients.
Easy-Medium
β€‹πŸ€–πŸ“•β€‹
2018
This course consists of half lectures and half paper reviews. Pascal Poupart does a phenomenal job in this course walking through theoretical concepts in a really approachable manner.
Medium
​
UCL & DeepMind - Intro to RL​
2015
This well-known RL course by David Silver goes into the nuts and bolts of RL from a time where Deep RL was still in its very infancy. The topics are explained more mathematically rigorously than in the UAlberta course, but lack the quizzes and coding practice that the MOOC offers.
Medium
β€‹πŸ“•β€‹
​
Supplementary Resources
Resource
Year
Description
Type
Barto & Sutton - RL: An Introduction​
2018
The textbook that is used in most RL courses. Written and explained clearly, this book is a great complement to any intro and more advanced RL course.
πŸ“š
OpenAI - Spinning Up​
2018
One of the best all-around deep RL resources out there. Whether you use the site as a starting point or review of key algorithms or reference for PyTorch/TensorFlow RL implementations, you’ll probably find it in their docs. There is also a complementary Spinning Up workshop recording worth watching if you want a quick intro to deep RL.
🌐
Arxiv Insights - YouTube channel​
Continously updated
Arxiv Insights is one of the best ML YouTube channels and has a lot of content on intro RL concepts.
πŸŽ₯
Lilian Wang - Lil'Log​
Continously updated
One of the best ML blogs and every post is a deep dive into another topic like attention, curricula or evolution strategies. A lot of posts are also about RL.
🌐
Last modified 5mo ago
Copy link