Mathematical Colloquium

Friday, November 23, 2018, 2:30pm

Newsvendor in Sobolev space, and beyond

Rüdiger Schultz (Uni Duisburg-Essen)

The newsvendor is among the classic problems in decision making under uncertainty. It consists of having to purchase for later use a commodity under uncertain demand. At demand realization, excess and shortfall of the commodity induce compensation costs; that add to the initial purchase cost. The task is to minimize total cost. In the talk we quickly depart from the newsvendor, extract basic features of two-stage stochastic programming, and visit more recent classes of stochastic programs. The first is PDE constrained optimization, where optimization of shapes under linearized elasticity and random loading is treated in detail. The second class of problems lives in finite-dimension. It is two-stage stochastic mixed-integer linear programs. For both classes we discuss structural properties beyond convexity, elaborate on solution methods, and report numerical testing for practical and academic applications

Location: Raum 008/SeMath, Pontdriesch 14-16, 52062 Aachen