Why one must use reweighting in Estimation Of Distribution Algorithms

Fabien Teytaud (1, 2) , Olivier Teytaud (3, 4, 5)

1 INRIA Futurs
2 UP11 UFR Sciences - Université Paris-Sud - Paris 11 - Faculté des Sciences
3 TANC - Algorithmic number theory for cryptology
4 LRI - Laboratoire de Recherche en Informatique
5 TAO - Machine Learning and Optimisation

We study the update of the distribution in Estimation of Distribution Algorithms, and show that a simple modification leads to unbiased estimates of the optimum. The simple modification (based on a proper reweighting of estimates) leads to a strongly improved behavior in front of premature convergence.

Domaines: Optimisation et contrôle [math.OC]

Fabien Teytaud, Olivier Teytaud. Why one must use reweighting in Estimation Of Distribution Algorithms. GECCO, 2009, Montréal, Canada. ⟨inria-00369780⟩