A computational evaluation of two subgradient search methods

Abstract The success of Lagrangean relaxation in discrete optimization is well documented. Part of this success is due to efficient procedures to find (near) optimal Lagrange multipliers, often by the use of subgradient search. This paper presents a comparison of the modified subgradient search technique of Camerini et al . and the most widely employed subgradient search procedure. Empirical results in different discrete optimization problems are given and discussed. Cases in which the modified subgradient search procedure is significantly superior and inferior are represented. Both procedures are easily implemented and thus should both be tested in special applications.