Generalized light robustness and the trade-off between robustness and nominal quality (Q471020)

Today when many researchers propose various algorithms for solving optimization problems it is very important to construct good algorithms. To be a good algorithm can mean also to give solutions applicable under uncertain circumstances. Thus, the purpose of robust optimization is to study the mentioned problem. The paper develops and considers a general notion of light robustness called generalized light robustness that is applicable to all types of optimization problems and to arbitrary uncertainty sets. Further this concept is compared to the known concepts in a number of theorems. Finally the trade -- off between robustness and nominal quality is analyzed emphasizing the possibility of a deviation from the optimal objective value in order to increase the robustness of the solution. The developed concept and analysis performed in this paper are very important since the purpose of new algorithms is not only to be new, but also to be applied in the practice. In order to apply new algorithms one has to be sure that they will work under many uncertainties.