Ôøú Ù Blockinøøóò× Äääöòòò Øó Ù×ø Øó Ö×

Abstra t: Au tion me hanism design has traditionally been a largely analyti pro ess, relying on assumptions su h as fully rational bidders. In pra ti e, however, bidders behave unpredi tably, making them dif ult to model and ompli ating the design pro ess. To address this hallenge, we present an adaptive au tion me hanism: one that learns to adjust its parameters in response to past empiri al bidder behavior so as to maximize an obje tive fun tion su h as au tioneer revenue. In this paper, we give an overview of our general approa h and then present an instantiation in a spe i au tion s enario. The algorithm is fully implemented and tested. Results indi ate that the adaptive me hanism is able to outperform any single xed me hanism.