Course Description

CSC2542 is a seminar-style topics course that explores recent advances in knowledge representation and automated reasoning. In the fall of 2016, the topic being covered is "Algorithms for Sequential Decision Making." Sequential decision making is the task of deciding what to do in the context of an extended interaction with the environment. It is a core competency of most intelligent agents. Central to sequential decision making is the notion of making a good decision now with respect to both immediate and longer term consequences. A driver deciding on a particular route to work, a robot deciding what to say to a person it is conversing with, a bank electing to embark on a major advertising campaign, or a computer program deciding on a move in a game of Go are all examples of sequential decision making.

This course will examine select principles and algorithmic techniques that are exploited for sequential decision making. The course will not be exhaustive but rather will focus on two themes: advances in techniques for search including suboptimal search algorithms, real-time search, and Monte Carlo search; and techniques for sequential decision making under uncertainty, including Markov Decision Processes (MDPs) and Reinforcement Learning (RL).

The course will draw predominantly on research readings. The format of the course will be a mix of class lectures, seminars, videos, assigned readings, and student paper presentations. A background assignment and a course project will be major components of a student's course mark.

Students taking the course should have the equivalent of an undergraduate introductory course in AI, such as CSC384, and reasonable competency as a programmer.

This should be a fun course! If you're thinking of taking the course and have questions, feel free to conact us.