On the similarities between AS, BSC and PBIL: toward the birth of a new meta-heuristic

One can notice that several work in genetic or ant-based optimization such as the BSC, PBIL or AS, are leading in the same direction. A new metaheuristic is about to appear, which looks for an optimum in a search space by using an explicit vector of probabilities that represents the actual sampling of the space. We introduce a common framework called PSM (Probabilistic Search Meta-heuristic) for describing such meta-heuristics, and we map BSC, PBIL and AS onto this framework. We then highlight the common points and diierences between these three algorithms. We provide an empirical comparison on standard binary functions where we test many diierent parameter values. We conclude on future research directions. 1 Au sujet des ressemblances entre AS, BSC et PBIL : vers la naissance d'une nouvelle m eta-heuristique Mots cl es optimisation, syst eme de fourmis, algorithme g en etique, syst eme a base de population