k-Distance Recoverable Robustness, seminar by Dr. Christina Büsing, RWTH Aachen University

Recoverable Robustness is a method to deal with uncertainties in an optimization problem and extends the classical concept of robustness by allowing limited recovery actions after all data are revealed. In this talk I will present a special case of this method where the recovery actions are limited by changing at most k elements of the previous fixed solution and apply this method to various combinatorial optimization problems. For the shortest path problem, we will see that small changes in the problem setting, e.g., choosing simple s,t-paths at the beginning instead of any s,t-path, have a big influence on the complexity status and combinatorial structures of the optimal solutions.

I will conclude the talk with an overview of current results on cutting planes for the recoverable robust knapsack polytope and an application to a train classification problem.

Most welcome!

Friday, March 15, 2013, 10:00 to 12:00
SICS, room Knuth
Isafjordsgatan 22
Markus Bohlin