ABSTRACT
This chapter presents basic learning schemes in games. It focuses on strategy learning in games under perfect monitoring of past actions. The chapter also focuses on learning schemes that does not require the monitoring assumption. It presents strategy learning algorithms for engineering games with action monitoring and one-step recall. The chapter provides the idea of reinforcement learning in the following: reinforcement learners interact with their environment and use their experience to choose or avoid certain actions based on their consequences. It describes the fully distributed learning schemes in which each player is assumed to follow a set of hybrid learning schemes but does not need to know the existence of a game. The chapter also provides partially- and fully-distributed learning strategies for finite games and dynamic games, respectively. It discusses how to combine various learning schemes based on mean field game dynamics.
