Kah, Christopher; Walzl, Markus - 2015
We propose a learning dynamic with agents using samples of past play to estimate the distribution of other players' strategy choices and best responding to this estimate. To account for noisy play, estimated distributions over other players' strategy choices have full support in the other...