Experimental studies

• 30 city problem, NC = 5000 cycles

• Q found to be (relatively) unimportant

GAs [Whitley, 89] found a tour of 424.635

Previous slide

Next slide

Back to first slide

View graphic version

Ant-density and ant-quantity have inferior results when compared to ant-cycle. This is due to the kind of feedback information which is used to direct the search process. Ant-cycle uses global information, that is, its ants lay an amount of pheromone which is proportional to the amount of trail whose ants whose tour was poor. On the other hand, ant-density and ant-quantity use only local information. The search is not directed by any measure of the final result achieved.

The optimal value of r= 0.5 can be explained by the fact that the algorithm, after using the greedy heuristic to guide search during the early stages of computation, starts exploiting the global information contained in the values delta Tij of trail. Ant-cycle clearly needs to be able to forget part of the experience gained in the past in order to better exploit new incoming global information.