CS 886

Advanced Topics in AI
Multiagent Systems

Instructor: Kate Larson
email: klarson@cs.uwaterloo.ca

Room: DC 3313
Schedule: Tuesdays and Thursdays 10:00-11:30

 

Schedule

Papers

Project Information

Announcements

Results from the tournament are found here!

On Tuesday, April 11 we will have a series of presentations on everyones projects. The goal of these presentations is to give others in the class a highlevel idea what you are working on. Each student (or group if applicable) has approoximately 5-7 minutes to talk. **There is no need to make slides.** Also, we will start later than normal. The class will start at **10:30** instead of 10:00.

Classes on March 24, April 4, and April 11 will be held in DC 3314.

Assignment 3 has been posted.
Assignment description
Source Code

I have updated the schedule. For the week of March 20th, we are having 6 presentations. You only need to submit reviews on two of the days.

Class for February 16th is cancelled, due to the University closure. On February 21 we will start with the class presentations. This means that we will skip the presentation on Multiagent Learning. I will be posting some notes and slides on single agent reinforcement learning so that students who have not done an AI course, will not be lost when we read the learning papers.
Notes on MDP's
Notes on single agent reinforcement learning


Assignment 2 is due on February 28th.
Here are some resources that you will find useful when doing Assignment 2. Assignment 2 has been posted.

Here is an example of a Bayes Nash equilibrium. (pdf).

Here is an example of a subgame perfect equilibrium in a bargaining game (pdf).

Assignment 1 can be found here. Office hours for CS 886 are on Tuesdays from 2:30-3:30.

Course Overview

The field of multiagent systems studies systems of multiple autonomous entities with diverging information and perhaps interests. This creates challenges above and beyond single-agent settings since we must now be additionally concerned with such issues as cooperation, coordination, and overcoming self-interest of individual agents in order to reach desirable system-wide goals. This course covers the mathematical and computational foundations of multiagent systems, with a focus on game theoretic analysis of systems in which agents can not be guaranteed to behave cooperatively.

Prerequisites

This course draws on a wide set of ideas from AI, CS theory and economics. While there are no formal prerequisites, some of the topics are quite formal mathematically, and students need to be able to construct and follow formal proofs.

Please send me email if you have any questions.

Course Topics (tentative list)

Organization

The course will be a combination of lectures and reading and discussion of research papers. Students will be given several homework assignments on the material covered in the lectures. With the research papers, students will be responsible for presenting them in class and discussing them. Projects will also be presented in class at the end of the semester.

Grading

Paper presentations
Assignments 
Class Participation
Term-long Project

15%
15%
15%
55%
 

Class Participation:

Class participation is an important component of this course. Before each class, all students must read the paper and submit comments and questions. Things to think about include

The Project:

The final project allows students to explore material not covered in class, and share that material with other students. The topic of the project can be a survey of a subarea of multiagent systems, a compare and contrast study of two or more influential papers, or a development of your own research ideas. Possible ideas for projects will be discussed in class.

The project will involve several steps

  • Project proposal
  • Project presentation
  • Final project report