CompSci 285: Multi-Agent Systems
CompSci 285: Multi-Agent Systems
Meeting: Tuesday/Thursday 10 - 11:30am Location: MD 123
First Meeting on 9/3
Instructor: David Parkes TF: Greg Stoddard
parkes -at- eecs.harvard.edu gstoddard -at- seas.harvard.edu
MD 229 MD 219
Office Hours: Office Hours: Wednesday 1:30 - 2:30
11:30 - 12:30 on 9/3, 9/5, 9/10
From the week of 9/17 on:
Tuesdays/Thursdays 2:30 - 4pm.
Beginning on 9/17, these Tue/Thur office hours are primarily to discuss readings with students presenting in class.
Announcements
- Here’s a list of potential ideas for course projects. Feel free to update it with your ideas.
- Submit comments for the reading using this form
-Please fill out this survey regarding class interest and experience
-We will be using Piazza for announcement and class discussions. https://piazza.com/harvard/fall2013/cs285/home
-There’s no reading for 9/3 (the first class) but there will be required reading for 9/5.
Course Description:
A multi-agent system is composed of multiple autonomous entities, with distributed information, computational ability, and possibly divergent interests. Multi-agent systems may be cooperative, such as sensor networks and mobile robots in a warehouse, or competitive, such as in electronic commerce, or in settings of resource or task allocation. Some systems are somewhere in between, for example Wikipedia and StackOverflow. The agents in a multi-agent system may be both artificial and human (and often humans are present at least on the edges of these systems.)
This is a seminar course on reading foundational and recent papers in various topics related to multi-agent systems, with a particular focus towards coordinating and promoting cooperative behaviors and an emphasis given to networks and the internet economy. The primary goal is to provide an introduction for students looking to identify research directions in this area. A secondary goal is to develop skills for critically reading research papers and for developing research questions.
General Topics
Algorithmic, game-theoretic and logical foundations of multi-agent systems, including distributed optimization and problem solving, non-cooperative game theory, learning and teaching, communication, social choice, mechanism design, auctions, negotiation, coalitional game theory, logics of knowledge and belief, collaborative plans and social systems
This year, an emphasis will be given to cooperative problem solving, cooperative game theory, axiomatic approaches, and social choice. The technical content will relate to machine learning and probabilistic reasoning, game theory, search, and preference elicitation. For specific topics, please see the class schedule.
Prerequisites
Computer Science 181 or 182, or permission of instructor. Familiarity with economic theory and computer science theory is helpful but not required. Students should be comfortable with formal, proof-based approaches.
Structure:
Throughout the course, we will be reading and discussing research papers. A strong emphasis is given to class participation.
After the first couple of weeks the discussion will be led by student presentations of material, likely in pairs (but depending on enrollment). Students will meet with the professor and TF to discuss the papers that they will present. All students are expected to submit comments on the reading to guide our in-class discussion.
There will be 2-3 relatively short problem sets, and these will comprise 20% of the grade. Participation in class discussion and through comments on papers, along with the presentation of research papers, will comprise 40% of the grade. As part of this, students will be expected to post a brief note on Piazza about a current interest topic related to things discussed in class.
Students will also be responsible for a research project or survey paper based off of a topic that we have discussed in class (or a topic that is closely related.) This will make up the remaining 40% of the grade.
Submitting Comments
Starting on Thursday 9/5, you must upload comments on the reading by midnight before class. Things to think about include (you don't need to hit all of these...):
what is the main contribution of the paper?
what was the main insight in getting the result?
what is not clear to you?
what are the most important assumptions, are they limiting?
what extensions does this suggest?
Projects
The goal of the final paper is to develop a deep understanding of a specific research area related to the topic of the class, and to the extent possible to work on an open research problem. Students are required to submit a proposal, give a short presentation, and submit a final paper at the end of reading week (maximum 10 pages except for Appendix material). Projects may be computational, theoretical, experimental or empirical. Students may also write an exposition paper on two related technical papers of their choice; expositions must cover two results from these papers, provide a critical discussion of assumptions made, and offer a new perspective.
Tuesday 11/12: Project proposals due
Tuesday 12/3 and Thursday 12/5: Project presentations
Monday 12/9: Final write-ups due at Noon
Collaboration Policy
For assignments, you are encouraged to consult with your classmates as you work on problem sets. However, after discussions with peers, make sure that you can work through the problem yourself and ensure that any answers you submit for evaluation are the result of your own efforts. In addition, you must cite any books, articles, websites, lectures, etc that have helped you with your work and list the names of students with whom you have collaborated on problem sets. For the final project, if you do a group project then everyone is expected to contribute effort and to fully understand the submitted paper.
Resources
There is no required text book. But the following provide useful materials and are on reserve in Pierce library:
Multiagent systems: Algorithmic, Game-theoretic and Logical Foundations, Shoham and Leyton-Brown CUP 2009
Algorithmic Game Theory, Nisan et al., CUP 2007
Multiagent Systems, G. Weiss (ed) MIT Press 2013
Networks, Crowds and Markets, D. Easley and J. Kleinberg, Cambridge University Press 2010
Related Courses
Below is a list of related courses that are being taught concurrently or in the following semester.
CS 181 (Spring 2014): An introduction to probabilistic reasoning and machine learning
CS 182 (Fall 2013): An introduction to artificial intelligence
CS 186 (Spring 2014): An introduction to topics at the intersection between economics and computer science.
CS 279 (Fall 2013): Research topics in Human-computer Interaction
CS 280r (Spring 2014): Advanced topics in artificial intelligence
CS 281 (Fall 2013): Advanced machine learning
CS 284r (Fall 2013): A research seminar related to social networks and game theory.
CS 286r (Spring and Fall 2014): Research seminar on Econ/CS topics-- expected Algorithmic Game Theory
(Spring 14) and Approximation in Mechanism Design (Fall 14).
CS 289 (Spring 2014): Biologically inspired Multi-agent systems