A Study on Sensitivity Analysis for Convex Quadratic Programs

Abstract

We extend the two similar interior-point approaches to sensitivity analysis originally developed for linear programs to those for convex quadratic programs, where the first approach is the ∊-sensitivity analysis and the other is Yildirim and Todd's. We study the relationship between the bounds on perturbation of the input parameters arising from the extension of Yildirim and Todd's approach and those from the ∊-sensitivity analysis. Furthermore, we prove that Yildirim and Todd's bounds are asymptotically the same as the symmetrized optimal tripartition bounds in the case of a special type of nondegeneracy.