Methodologies: Machine Learning, Uncertainty Quantification

Michael Wellman

Professor, Electrical Engineering and Computer Science, CSE


Center for the Study of Complex Systems

In a strategic environment, agents face decisions where the outcomes depend on the behavior of other autonomous agents. The strategic reasoning group develops techniques for understanding and engineering complex multiagent environments, using concepts and methods from economics as well as computer science. Specifically, we apply game-theoretic principles to data from large-scale agent-based simulation, in an approach called empirical game-theoretic analysis (EGTA). EGTA combines simulation, machine learning, and other empirical methods to reason about the strategic issues in complex multiagent settings. We are particularly interested in domains characterized by dynamism, networks, and uncertainty, including applications in financial markets, information security, and sustainable transportation.

The iterative empirical game-theoretic analysis process.

The iterative empirical game-theoretic analysis process.