The Small World of Pheromone Trails
In this paper we consider \(\cal M\!AX\!\)–\(\cal MI\!N\!\) Ant System and Ant Colony System. They are generally recognized to be the best performing algorithms of the Ant Colony Optimization family. They are characterized by a quite different way for dealing with the pheromone trail. We propose an experimental analysis for observing whether this difference impacts significantly on the characteristics of the pheromone distributions produced during the runs. The results obtained are analyzed by using some concepts derived by the literature on small-world networks. It comes out that ants actually build small-world pheromone graphs during their runs. This behavior is interpreted here as a sort of decomposition of the instances tackled.
KeywordsTravel Salesman Problem Small World Pheromone Trail Objective Function Evaluation Characteristic Path Length
Unable to display preview. Download preview PDF.
- 2.Pool, I., Kochen, M.: Contacts and influence. Social Networks 1, 1–48 (1978)Google Scholar
- 3.Milgram, S.: The small world problem. Psychology Today 2, 60–67 (1967)Google Scholar
- 4.Newman, M.E.J., Barabasi, A.L., WattsD., J.: The Structure and Dynamics of Complex Networks. Princeton University Press, Princeton (2006)Google Scholar
- 5.Stützle, T., Hoos, H.H.: Improving the Ant System: A detailed report on the MAX–MIN Ant System. Technical Report AIDA-96-12, FG Intellektik, FB Informatik, Technische Universität Darmstadt, Darmstadt, Germany (1996)Google Scholar
- 8.Bollobas, B.: Random Graphs. Academic Press, London (1985)Google Scholar
- 13.Birattari, M.: The Problem of Tuning Metaheuristics as Seen from a Machine Learning Perspective. PhD thesis, Université Libre de Bruxelles, Brussels, Belgium (2004)Google Scholar
- 14.Pellegrini, P.: ACO: parameters, exploration and quality of solutions. PhD thesis, Department of Applied Mathematics, Università Ca’ Foscari, Venice, Italy (2007)Google Scholar
- 15.Friedman, J.H.: A variable span smoother. Technical Report 5, Department of Statistics, Stanford University, Stanford, CA, USA (1984)Google Scholar