On the Invariance of Ant System
It is often believed that the performance of ant system, and in general of ant colony optimization algorithms, depends somehow on the scale of the problem instance at hand. The issue has been recently raised explicitly  and the hyper-cube framework has been proposed to eliminate this supposed dependency.
In this paper, we show that although the internal state of ant system—that is, the pheromone matrix—depends on the scale of the problem instance under analysis, this does not affect the external behavior of the algorithm. In other words, for an appropriate initialization of the pheromone, the sequence of solutions obtained by ant system does not depend on the scale of the instance.
As a second contribution, the paper introduces a straightforward variant of ant system in which also the pheromone matrix is independent of the scale of the problem instance under analysis.
Unable to display preview. Download preview PDF.
- 3.Dorigo, M., Maniezzo, V., Colorni, A.: The Ant System: An autocatalytic optimizing process. Technical Report 91-016 Revised, Dipartimento di Elettronica, Politecnico di Milano, Milano, Italy (1991)Google Scholar
- 4.Dorigo, M.: Ottimizzazione, apprendimento automatico, ed algoritmi basati su metafora naturale. PhD thesis, Politecnico di Milano, Milano, Italy (1992) (in Italian)Google Scholar
- 6.Birattari, M., Pellegrini, P., Dorigo, M.: On the invariance of ant colony optimization. Technical Report TR/IRIDIA/2006-004, IRIDIA, Université Libre de Bruxelles, Brussels, Belgium (2006) (Submitted for journal publication)Google Scholar
- 7.Stützle, T., Hoos, H.H.: The \(\cal MAX\)–\(\cal MIN\) Ant System and local search for the traveling salesman problem. In: Bäck, T., Michalewicz, Z., Yao, X. (eds.) Proceedings of the 1997 IEEE International Conference on Evolutionary Computation (ICEC 1997), Piscataway, NJ, USA, pp. 309–314. IEEE Press, Los Alamitos (1997)CrossRefGoogle Scholar