Algorithms for Computing Set Tolerances: Theory, Computational Analysis and Applications

Abstract

The regular set tolerance is an important term in sensitivity analysis. For combinatorial sum problems, e.g., the Traveling Salesman Problem, Shortest Path Problem and Minimum Spanning Tree Problem, it determines how much the sum of the costs of the elements of a set can be increased while ensuring that all current optimal solutions remain optimal. The regular set lower tolerance determines how much the sum of the costs of the elements of a set can be decreased while ensuring that the objective value of the optimal solution is not changed. We investigate general methods for computing regular (upper and lower) set tolerances in combinatorial sum problems. For the upper tolerance, we present a linear programming approach, and for the lower tolerances, three linear programming approaches, where the last two are novel and lead to recursive procedures for computation of the lower tolerances of all subsets of the given ground set. We give new upper bounds for set lower tolerances. For both upper and lower tolerances, we give an exact formula for sets of cardinality 2 and 3. Furthermore, we computationally compare and analyze the three different lower tolerance LPs. Finally, we consider the computation of tolerances for the Minimum Spanning Tree Problem, give a formula for single tolerances, a lower bound for regular set upper tolerances and an exact formula for regular set lower tolerances.