Spring 2013

CSCI 8920 - Decision Making Under Uncertainty

Instructor: Prashant Doshi

Class times: Tue, Thurs 11:15p - 12:05p

                         Wed 11:00p - 12:15p

Choosing optimally among different lines of actions is a key aspect of autonomy in artificial agents. The process by which an agent arrives at this choice is complex, particularly in environments that are noisy and/or shared with other agents. This course will focus on how to make optimal and approximately optimal decisions in multiagent settings. It will be self-contained, introducing relevant background literature such as aspects of probability and game theories. A tentative list of topics covered in the course is given below:

I. Introduction

-Requirements for decision models and solutions

-Probability theory background

-Bayesian networks and Influence diagrams

II. Decision making in single agent setting

-Markov decision processes (MDP)

-Partially observable Markov decision processes (POMDP)

-Dynamic influence diagrams

III. Decision making in multiagent setting

-Game theory background

-Repeated strategic and Bayesian games

-Decentralized MDPs

-Partially observable Dec-MDPs (DEC-POMDPs) and approximations

-Interactive POMDPs (I-POMDPs) and approximations

-Interactive Influence diagrams

The course will adopt a unique pedagogical style, utilizing some classroom games to generate intuition and reinforce instruction and presentations of research papers by students toward the end of the course. Please contact the instructor at pdoshi@cs.uga.edu for further questions.