ON LOWER BOUND UPDATES IN PRIMAL POTENTIAL REDUCTION METHODS FOR LINEAR-PROGRAMMING

We present a procedure for computing lower bounds for the optimal cost in a linear programming problem. Whenever the procedure succeeds, it finds a dual feasible slack and the associated duality gap. Although no projective transformations or problem restatements are used, the method coincides with the procedures by Todd and Burrell and by de Ghellinck and Vial when these procedures are applicable. The procedure applies directly to affine potential reduction algorithms, and improves on existent techniques for finding lower bounds.