Teaching

Part of the Data Mining and Machine Learning (DMML) course this 8 lecture series introduces Reinforcement Learning. It covers the basics of Makovian states, V* values, Q functions and Q-learning based on Chapter 13 of Machine Learning by Tom M. Mitchell (1997 edition, publishers: McGraw-Hill). It also draws on material from Reinforcement Learning an Introduction by Richard S. Sutton and Andrew G. Barto (1998, MIT Press), specifically n-step TD updates, eligibility traces, function approximation, direct policy search and Actor-Critic methods.

The RL assignment (Assignment 2) is now available here.

I strongly suggest reading Sections 13.1 to 13.4 of Machine Learning and the introductory sections (Sections 1.1-1.3) of Reinforcement Learning an Introduction.

Copies of the 2011-12 slides (uploaded after each lecture) and some demonstration videos can be found below. For those who missed my introduction to confidence intervals the following YouTube video may be of help. http://www.youtube.com/watch?v=Hn6C21GC0vA