CS 6804: Learning and Sequential Decision-Making
Dr. Sanmay Das
Spring 2013

This course will study, from a computer science perspective, the
problems faced by agents who are situated in dynamic, changing
environments. The central question is how agents should reason in
order to make the best decisions in these environments. In order to do
so, they must be able to learn from experience and perform complex
decision-making tasks under uncertainty. Topics to be covered include,
but are not limited to, Markov decision processes, partial
observability, reinforcement learning, bandit problems, sequential
search, and reasoning and learning in games.

The course will largely follow a seminar format. Students will be
expected to complete the assigned reading before class and come to
class ready to engage in discussing the reading. Most of the reading
list will be drawn from recent literature (such as the ICML, NIPS,
AAAI, IJCAI, AAMAS and EC conferences). The major deliverable will be
a group project that students will work on through the course of the
semester.

Prerequisites: Undergraduate class in AI (CS 4804 or equivalent) or
machine learning or data analytics. Graduate work in
algorithms/machine learning/data mining strongly
recommended. Familiarity with probability and ability to program
(ideally in a language that is mathematically flexible and well-suited
for rapid prototyping, e.g. Matlab, R, or Python) are absolutely
essential.