A Grasp for Qap 214

Du sequentiel au parallele: la recherche aborescente a la programmation quadra-tique en variables 0 et 1, 1987. Th ese d'Etat. 33] , A parallel branch and bound algorithm for the quadratic assignment problem, Discrete A GRASP FOR QAP 21 4. Concluding remarks. In this paper, we discussed aspects of a GRASP implementation for solving the QAP. The algorithm was tested on a broad range of test problems and produced good-quality solutions in a reasonable amount of CPU time. Best known solutions were produced for almost all of the instances tested. In a few cases, the permutations found were better than those previously reported in the literature. The algorithm can be easily implemented on a parallel computer since diierent GRASP iterations can be assigned to each processor. A single global variable (value of best solution found) is shared by the processors. It is expected that a parallel implementation should improve the computational results given in this paper as was observed in 14]. GRASP can be also adapted to solve special cases of QAP, like the traveling salesman problem, graph isomorphism, graph partitioning, and the band-width reduction problem. A greedy randomized adaptive search procedure for the maximum independent set, Problem Optimal Best value GRASP GRASP CPU time Name value found iterations local total lipa10a 473 473 4 0.01 0.01 lipa20a 3683 3683 35 0.86