Revisiting log-linear learning: Asynchrony, completeness and payoff-based implementation

The theory of learning in games has sought to understand how and why equilibria emerge in non-cooperative games. Traditionally, social science literature develops descriptive game theoretic models for players, analyzes the limiting behavior, and generalizes the results for larger classes of games. Recently, there has been a significant amount of research seeking to understand these behavioral models not from a descriptive point of view, but rather from a prescriptive point of view [1]-[4]. The goal is to use these behavioral models as a prescriptive control approach in distributed multi-agent systems where the guaranteed limiting behavior would represent a desirable operating condition.