Publication at Faculty of Mathematics and Physics |
2008
Abstract
We consider a linear programming problem and develop an effective method for computing tolerances for input data. The tolerances are determined such that the input quantities can simultaneously and independently vary within these tolerances while the optimal value does not exceed given lower and upper bounds.
In our approach we are able to take into account all the input quantities or some selected ones. The procedure runs in polynomial time.
Although the tolerances are not the best possible (due to dependencies between quantities) in general, the results are satisfactory. We illustrate the procedure on a simple portfolio selection problem modelled as a linear program.