CPS 173: Computational Microeconomics, Spring 2010

WF 10:05-11:20am, North Building 306.
Instructor: Vincent Conitzer. (Please call me Vince.)

Teaching Assistant: Vincent Conitzer.

In recent years, there has been a surge of interaction between computer scientists and economists. This interaction is driven both by necessity and opportunity. On the one hand, as computer systems become more interconnected, multiple parties must interact in the same environment and compete for scarce resources, which necessarily introduces economic phenomena. On the other hand, in the past, economic mechanisms (such as auctions and exchanges) have been designed to require very limited computing and communication resources, as they would otherwise be impractical. These days, computation and communication pose much less of a constraint, which presents an opportunity to create highly efficient, computationally intensive mechanisms.

In the first part of the course, we will study the design of expressive marketplaces. In such marketplaces, participant can express nontrivial valuations over outcomes: for example, a participant may express that a complete travel package to Las Vegas including a flight, hotel reservation, and entertainment is worth $700 to her, but any incomplete package is worth $0. This can greatly increase market efficiency, but clearing the market (deciding on the final outcome) becomes computationally hard. We will cover techniques for solving these problems.

In the second part of the course, we will study game theory. Game theory studies how to act optimally in strategic settings where each party's utility (happiness) depends on the actions of other parties. We will cover such definitions of optimality as well as techniques for computing optimal actions. We will study applications including bidding in auctions and building computer poker players.

In the third part of the course, we will draw on the first two parts and study how to design market mechanisms that are optimal when we take into account that agents will behave strategically (game-theoretically). Again, we will cover techniques for computing the optimal mechanisms.

Students should be comfortable with probability. Background in computer science and/or economics will be helpful but neither is required; the goal is to bring together students from different backgrounds. While there are no other specific mathematical prerequisites, the course will probably not be enjoyable for students who dislike mathematics.

We will use parts of a new book by Shoham and Leyton-Brown (SLB), Multiagent Systems. A free electronic copy is available at that link though the printed version is very reasonably priced as well.
There will be additional readings for individual classes. The slides for the course are also part of the reading.

Grading (subject to change)
Participation: 10%
Programming assignments: 15%
Written assignments: 15%
Midterm: 15%
Small project: 20%
Final exam: 25%

Rules for assignments: You may discuss homework assignments with at most one other person, in person. However, you may not simply copy down the other person's solution (or any part thereof). Each person should do her/his own writeup, at which point you should derive the solution yourself. This also implies that you cannot copy any code (linear programs etc.) from each other. Copying code is considered a serious form of cheating, and there are ways of detecting copied code. If you have trouble with the programming assignments, just ask for help. On your writeup, you should acknowledge your partner (if any) and any other sources you used.

Rules for exams: Exams will be closed-book. However, you do not need to remember every detail of the modeling language (where the colons go, for example).

Rules for the project: You may work on the project in a team of size 1, 2, or 3. If there is a good reason to have even more people on the team, discuss this with Vince. Unlike the assignments, in the project you are allowed to share everything, including code, within the project team.
The goal for the project is to allow you to get creative with some of the material in the course. For example, find an interesting problem related to computational microeconomics and discuss how to solve it using LP/MIP (or other techniques). If you have trouble thinking of topics, Vince can help you, but try to come up with your own. More specific guidelines can be found here.
Since this is the first time I am (or anyone else is) teaching this course, we will not plan the course down to the individual lecture. Dates will be added as the course progresses. Topics are given below (a topic need not take exactly one lecture to complete and we may not cover all topics).

Date Topic Materials
1/15 Course at a glance. Slides: ppt, pdf.
Optional: CACM overview article.
Part 0: Basic techniques from computer science.
1/20 - 1/29 Linear programming. (Mixed) integer linear programming. Slides: ppt, pdf.
Example files: class_example.lp, class_example2.lp, class_example3.lp, class_example4.lp, knapsack.lp, knapsack_simple.mod, knapsack.mod, cell.lp, cell.mod, hotdog.mod, sudoku.mod.
SLB Appendices A, B.
Programming assignment 1 out (due 2/3).
Guide to the modeling language.
2/3, 2/5 Computational problems. Algorithms. Runtime of algorithms. Easy and hard problems. Slides: ppt, pdf.
Sorting algorithms spreadsheet.
Example files: set_cover.mod, set_cover2.mod, matching.mod.
Optional: CACM article on P vs. NP.
Part 1: Expressive marketplaces.
2/10 - 2/19 Single-item auctions. Combinatorial auctions. Bidding languages. Winner determination problem. Variants (reverse auctions, exchanges). Slides: ppt, pdf.
Note: we are not going in the same order as the book in these lectures. The book does mechanism design before getting into auctions.
SLB 11.3.1-11.3.4, 11.4.1.
Optional: 11.2, 11.3.5, Conitzer chapter on auctions, Lehmann et al. chapter on winner determination, Sandholm chapter on optimal winner determination.
2/24, 2/26 Expressive financial securities. Slides: ppt, pdf.
SLB 10.4.2.
Programming assignment 2 out. Partial solution to graph winner determination problem, for first problem.
Optional: Paper 1, paper 2.
3/3, 3/5 Barter exchanges/matching markets. Kidney exchange. Slides: ppt, pdf.
Paper (you do not need to understand all the details about constraint and column generation).
3/17, 3/24 Voting and social choice. Slides: ppt, pdf.
SLB Chapter 9 (9.5 is optional).
Part 2: Game theory.
3/26 Risk neutrality and risk aversion. Expected utility theory. Slides: ppt, pdf.
SLB Section 3.1.
3/26, 3/31, 4/2, 4/7 Games in normal form. Dominance and iterated dominance. Computing dominated strategies. Minimax strategies. Computing minimax strategies. Nash equilibrium. Computing Nash equilibria. Homework 3 out.
Slides: ppt, pdf.
SLB 3.2, 3.4.3, 4.5; 3.3.1-3.3.3, 3.4.1, 4.1, 4.2.1, 4.2.3, 4.2.4, 4.4.
Optional (including the papers): 3.3.4, 4.2.2; 3.4.5, 4.6. Paper on computing dominated strategies. (You can skip the part on Bayesian games.) Paper on computing Nash equilibria. (You only need to read the part concerning 2-player games.) Paper on computing special kinds of Nash equilibria. (You can skip everything from Bayesian games on.)
4/9, 4/14 Games in extensive form. Backwards induction. Subgame perfect equilibrium. Imperfect information. Equilibrium refinements. Computing equilibria. Poker. Slides: ppt, pdf.
SLB 5.1 (alpha-beta is optional), 5.2.1, 5.2.2.
Optional (including the paper): 5.2.3. Paper on finding optimal strategies to commit to.
Part 3: Mechanism design.
4/16, 4/21 Bayesian games. Auctions revisited. Homework 4 out.
Slides: ppt, pdf.
SLB 6.3, 11.1.1-11.1.8.
Optional: 11.1.9, 11.1.10.
4/23 Incentive compatibility. Individual rationality. Revelation principle. Clarke mechanism. Groves mechanisms. Slides: ppt, pdf.
SLB 10.1-10.4.
Optional: rest of chapter 10.
4/28 Real-world applications.
Tuesday May 4, 7pm-10pm Final exam.