Learning Dynamics and Equilibrium Computation in Games

This theme studies learning algorithms for computing equilibria in games. I focus on algorithms whose strategies converge to Nash equilibrium in a last-iterate sense, including perturbation-based methods for stabilizing learning dynamics.

Key questions

#

• How can learning algorithms converge quickly to Nash equilibrium?
• How can perturbation-based methods stabilize learning dynamics?
• When do multi-agent learning dynamics fail to converge?