Abstract
The optimization function of the method of search-space smoothing in combination with the Metropolis algorithm (MASSS) is studied on the random Euclidean traveling salesman problem. Three numerical experiments conducted for the method of simulated annealing (SA) are performed to compare functional characteristics between these two methods. First, the optimization characteristics of MASSS are analyzed by using adaptive de-smoothing schedule, and the results show the existence of the effective value of the smoothing factor. Second, interbasin transition dynamics are investigated to evaluate their intensification function. The effect of the relaxation dynamics is maximized at some intermediate value of the smoothing factor, and the value is very close to that found in the first experiment. Last, rate-cycling experiment is performed to test directly the role of the relaxation dynamics in the de-smoothing process. In the optimization process of MASSS, effective relaxation dynamics and the resulting performance is sensitive to the search around the effective value of the smoothing factor. These functional characteristics are quite similar to those of SA.
