Abstract
We consider a general class of problems where a mechanism designer must elicit private information held by agents in order to maximize revenue. Examples of this are combinatorial auctions, multiproduct monopoly problems, market regulation in the presence of collusion, etc. By now, there is a standard way to formulate such a problem, but an analytical solution is unlikely to exist for agents with multidimensional private information.
In our work, we build on an approximating strategy pioneered by Ekeland and Moreno (2008) to solve a discrete version of the problem. By using a constraint generation approach, we solve big instances of the problem by using less than 1% of the constraints. Using convexification algorithms, an optimality bound can be calculated in each iteration, allowing for significantly faster solutions if a very small tolerance of optimality is allowed.

We provide numerical solutions for the multiproduct monopolist problem and discuss its implications.

13:45-14:15 hrs. Nicolás Figueroa. Profesor Instructor del Departamento de Ingeniería Industrial, Universidad de Chile. Ph. D en Economí¬a - Universidad de Minnesota, EE.UU.

A partir de las 13:30 hrs. quienes hayan confirmado asistencia podrán compartir de un grato almuerzo.