Learning payoff functions in infinite games (Q2384147)

The authors consider a class of games with real-valued strategies and payoff information available only in the form of data from a given sample of strategy profiles. Solving such games with respect to the underlying strategy space requires generalizing from the data to a complete payoff-function representation. They address payoff-function learning as a standard regression problem, with provision for capturing known structure (e.g., symmetry) in the multiagent environment. To measure learning performance, they consider the relative utility of prescribed strategy, rather than the accuracy of payoff functions per se. They check their approach on two examples: a two-player version of the first-price sealed-bid auction (with known analytical form), and a five-player market-based scheduling game (with no known solution). Additionally, they explore the efficacy of using relative utility of strategies as a target of supervised learning and as a learning model selector. Their experiments demonstrate effectiveness in the former case, though not in the latter.