Game Theory: Learning in Finite Action Games

This lecture discusses the learning dynamics in finite action games, focusing on the importance of game theoretic analysis. It begins with a motivation for understanding the problem and how game theory can provide solutions. The instructor outlines the agenda, which includes exploring different types of equilibria such as correlated and coarse correlated equilibria. The lecture recaps best-response dynamics and their convergence properties, emphasizing the challenges in finding pure and mixed Nash equilibria. An example of a Stop & Go game illustrates the concepts, highlighting the utilities and potential outcomes for players. The discussion extends to correlated equilibria, explaining how they differ from Nash equilibria and their implications for social welfare. The instructor also introduces coarse correlated equilibria, providing definitions and examples to clarify the concepts. The lecture concludes with a focus on computational methods for finding these equilibria, emphasizing their tractability compared to traditional Nash equilibria, and sets the stage for further exploration of learning algorithms in game theory.