Parallel CLM starting solutions comparison*

Resumen

Local search methods are sensitive to the way initial solutions are generated. In this communication the differences in terms of quality of solutions and execution time due to the method used to generate starting solutions are shown. Guided initial solutions usually lead to higher quality final solutions but the algorithm can converge quickly to poor quality local optima. Using randomly generated starting solutions could supply more diversified solutions at a cost of a greater computation time. In this communication, both methods are compared when applied in a parallel algorithm to the permutation flow shop problem. Keywords: parallel algorithms, local search