Saturation in linear optimization

Abstract

In a solvable linear optimization problem, a constraint is saturated if it is binding at a certain optimal solution and it is weakly saturated if it is binding at a proper subset of the optimal set. Nonsaturation and weak saturation can be seen as redundancy phenomena in the sense that the elimination of a finite number of these constraints preserves the value of the given problem. We consider also the effect of sufficiently small perturbations of the cost coefficients in the classification of a given constraint as either saturated or nonsaturated.