There is no required textbook for this course, but the following are the most related books:
In many real-world problems, we wish for a system to obtain information from self-interested participants (agents) and then decide an outcome that maximizes some notion of utility. Auctions and machine learning algorithms are just a few examples of such systems. The agents being self-interested poses a key challenge: the mapping from the agents' reported information to the outcome (which could involve giving partipants some goods, payments, or both) must be carefully selected in order to incentivize partipants to report their information truthfully. Mechanism design, a subfield of algorithmic game theory, is about the design of such systems. Designing a system can be thought of as designing the rules of the game. How can we design a mechanism to achieve favorable outcomes? What can go wrong if a mechanism isn't designed well?
A major theme of this course is the interaction between economics and computation. In particular, we will look at connections between learning algorithms and mechanism design. Some of these connections include: how to use learning algorithms to approximate various notions of equilibria (essentially, stable strategies) of games; how to learn classifiers in the presence of strategic agents that provide the data points; how to sequentially learn from self-interested experts while simultaneously incentivizing the experts to truthfully report their information.
This course is an advanced topics course, intended for research-focused graduate students as well as advanced undergraduate students. In the first major part of the course of the course, the instructor will give lectures on foundational material in mechanism design and online learning. This foundation will set the stage for us to focus on research papers. In the second major part of the course, in each lecture a pair of students or one student (or occasionally the instructor, depending on enrollment) will present a research paper to the class and lead an in-class discussion of that paper.
In the schedule below:
|