|Professor Amy Sliva|
|Office: 256 West Village H|
|Office hours: W 10:00-11:00am, F 1:30-2:30pm, or by appt.|
In many domains---game playing, economics, politics, etc.---it is
necessary to make decisions strategically, that is, based on
the predicted behavior of someone else. For example, when determining your next
move in chess, you must consider the move your opponent just made, as
well as what he is likely to do in future turns. In many
circumstances, however, making these strategic decisions is extremely
difficult (even impossible) for human analysts due to the size,
complexity, or uncertainty, of the problem domain.
The field of artificial intelligence has developed many techniques to automate this process of behavioral analysis and decison-making with special attention paid to situations of high uncertainty. This course will survey these methods, providing a look at tools used to model behavior and make strategic decisions. The course will be divided into two main units: 1) behavior representation and modeling, and 2) decision-making under uncertainty. We will study both the basic techniques and algorithms related to topics such as state-space search, logic, constraint satisfaction, and probabilistic graphical models, as well as examine new research that utilizes these tools for predictive modeling and decision-making.
Prerequisites: This course assumes a basic familiarity with search algorithms (i.e., depth first, breadth first, heuristic, etc.), propositional and first-order logic, probability theory, and basic complexity theory.