Incentives in dynamic markets

Abstract

In this thesis, we consider a variety of combinatorial optimization problems within a common theme of uncertainty and selfish behavior. In our first scenario, the input is collected from selfish players. Here, we study extensions of the so-called smoothness framework for mechanisms, a very useful technique for bounding the inefficiency of equilibria, to the cases of varying mechanism availability and participation of risk-averse players. In both of these cases, our main results are general theorems for the class of (lambda,mu)-smooth mechanisms. We show that these mechanisms guarantee at most a (small) constant factor performance loss in the extended settings. In our second scenario, we do not have access to the exact numerical input. Within this context, we explore combinatorial extensions of the well-known secretary problem under the assumption that the incoming elements only reveal their ordinal position within the set of previously arrived elements. We first observe that many existing algorithms for special matroid structures maintain their competitive ratio in the ordinal model. In contrast, we provide a lower bound for algorithms that are oblivious to the matroid structure. Finally, we design new algorithms that obtain constant competitive ratios for a variety of combinatorial problems.