Learning algorithms where each agent only observes its own payoffs and actions, not other players' strategies.