Reinforcement Learning

(Part of COMPGI13/COMPM050, along with Kernel Methods by Arthur Gretton)

This page refers to the previous 2016 course at UCL.

See the UCL Moodle for slides from the current course.

Meeting Time: Thursdays 9:15 am in Roberts 421 (@ University College London)

Lecturers:

Joseph Modayil Course website (modayil@google.com)

Hado van Hasselt Course website (hado@google.com)

Teaching Assistant:

Zbigniew Wojna (zbigniewwojna@gmail.com)

Assignment: Easy21 (posted: Feb 26, due: April 6)

Reference Text: Reinforcement Learning (draft 2nd edition) by Sutton and Barto

Reference Text: Algorithms for Reinforcement Learning by Csaba Szepesvari

Lecture Slides (with thanks to Dave Silver)

Lecture 1: Introduction to Reinforcement Learning (slides: Jan 14)

Lecture 2: Exploration and Exploitation in Bandits (slides: Feb 10)

Lecture 3: Markov Decision Processes (slides: Jan 27)

Lecture 4: Planning by Dynamic Programming (revised slides: Feb 6)

Lecture 5: Model-Free Prediction (slides: Feb 11)

Lecture 6: Model-Free Control (slides: Feb 25)

Lecture 7: Value Function Approximation (slides: Mar 11)

Lecture 8: Policy Gradient Methods (slides: Mar 10)

Lecture 9: Integrating Learning and Planning (slides: Mar 16)

Lecture 10: Case Study: RL in Classic Games (Regular lecture in class at 9:15am and guest lecture by David Silver at 1pm in Roberts G06)