An approach to calmness of linear inequality systems from Farkas lemma

Abstract

We deal with the feasible set mapping of linear inequality systems under right-hand side perturbations. From a version of Farkas lemma for difference of convex functions, we derive an operative relationship between calmness constants for this mapping at a nominal solution and associated neighborhoods where such constants work. We also provide illustrative examples where this approach allows us to compute the sharp Hoffman constant at the nominal system.