Dissecting “Reinforcement Studying” by Richard S. Sutton with Customized Python Implementations, Episode I
Reinforcement Studying (RL) is a captivating subfield of Machine Studying. You may already realize it from functions akin to enjoying Go [1], autonomous driving [2], and extra.
Equally fascinating in my view is Sutton’s and Barto’s well-known e book, “Reinforcement Studying” [3]. I believe it’s an ideal introduction to the subject, but additionally dives deep and introduces all essential theoretical subjects of the sector. It may be rather a lot to learn although, and particularly upon the primary learn may look a bit mathy.
Thus, I made a decision to begin a publish sequence summarizing the e book chapter by chapter. I consider getting the contents defined with totally different phrases will enormously assist understanding. And I will even implement all (most) algorithms from the e book in Python and apply them to issues and environments modeled by way of (previously) OpenAI’s gymnasium framework [4]. These two factors are, so far as I do know, novel to date and make this sequence distinctive.
This publish is the primary within the sequence, and can briefly introduce RL usually, then give a fast overview of how Sutton’s e book is structured — and the way…