The possible and the impossible in multi-agent learning

This paper surveys recent work on learning in games and delineates the boundary between forms of learning that lead to Nash equilibrium, and forms that lead to weaker notions of equilibrium or to none at all.