Network Reoptimization Algorithms: A Statistically Designed Comparison

Statistical design and analysis of experiments is a cornerstone methodology for scientific exploration and verification that is rarely employed in reporting mathematical software testing. This approach is used to study the relative effectiveness of three key algorithms for the reoptimization of network flow problems primal simplex, dual method, and out-of-kilter. Carefully designed experimentation with state-of-the-art codes accompanied by a rigorous statistical analysis isolates the most efficient method(s) for commonly encountered reoptimization problems, and identifies the effect on solution time of problem class, problem size, type of change (bounds, costs, or node requirements), degree and number of changes in parameters, and number of problems in the reoptimization series.

INFORMS Journal on Computing, ISSN 1091-9856, was published as ORSA Journal on Computing from 1989 to 1995 under ISSN 0899-1499.