Abstract: Mechanism design studies situations where a set of agents each hold private information regarding their preferences over different outcomes. A mechanism receives claims about preferences and chooses an outcome and payments. The typical approach is to impose incentive-compatibility (IC), and derive an optimal mechanism subject to this constraint. By replacing strict IC with the approach of minimizing expected ex post regret, we can adapt statistical machine learning to design suitable payment rules. Given a target outcome rule, we learn a payment rule with suitable properties via the discriminant function of a multi-class classifier. Applications are demonstrated to welfare-maximizing combinatorial auctions (with approximate and optimal clearing algorithms), as well as the egalitarian assignment problem. Based on "Payment Rules through Discriminant-Based Classifiers", Duetting et al. ACM Trans. on Economics and Computation 3(1), 5:1, 2015