Boulder Data Science, Machine Learning & AI

Last meeting we finished Chapter 4 on Dynamic Programming. For this meeting, we will start Chapter 5 on Monte Carlo methods for evaluation and control and try to get through section 4.4 and exercise 4.4. "On Policy" evaluation and control are the subjects of this part of the chapter which means that we are generating samples from an agent interacting with an environment following the policy of interest. We can also discuss the Monte Carlo sampling approach and how it differs from Dynamic Programming in terms of efficiency and convergence.

Useful Links:

Reinforcement Learning: An Introduction by Richard S. Sutton and Andrew G. Barto. The full book and supplemental material is available here: http://incompleteideas.net/book/the-book.html

My exercise solutions and chapter notes:
https://github.com/jekyllstein/Reinforcement-Learning-Sutton-Barto-Exercise-Solutions

Kickoff Slides which contain other links: https://docs.google.com/presentation/d/1QD3iw5BgIpPpl_K_ApAlDr1NRseR1WmXme1dKQGqTOg/edit?usp=sharing

Video lectures from a similar course: https://youtube.com/playlist?list=PLqYmG7hTraZDVH599EItlEWsUOsJbAodm