Sustainable Cooperative Coevolution with a Multi-Armed Bandit

François-Michel de Rainville (1) , Michèle Sebag (2, 3) , Christian Gagné (1) , Marc Schoenauer (3, 2) , Denis Laurendeau (1)

1 LVSN - Laboratoire de Vision et Systèmes Numériques
2 LRI - Laboratoire de Recherche en Informatique
3 TAO - Machine Learning and Optimisation

This paper proposes a self-adaptation mechanism to manage the resources allocated to the different species comprising a cooperative coevolutionary algorithm. The proposed approach re-lies on a dynamic extension to the well-known multi-armed bandit framework. At each iteration, the dynamic multi-armed bandit makes a decision on which species to evolve for a generation, using the history of progress made by the different species to guide the decisions. We show experimentally, on a benchmark and a real-world problem, that evolving the different popula-tions at different paces allows not only to identify solutions more rapidly, but also improves the capacity of cooperative coevolution to solve more complex problems.

Mots clés: Cooperative Coevolution; Multi-Armed Bandit; Credit Assignment; Adaptivity

Domaines: Intelligence artificielle [cs.AI]

François-Michel de Rainville, Michèle Sebag, Christian Gagné, Marc Schoenauer, Denis Laurendeau. Sustainable Cooperative Coevolution with a Multi-Armed Bandit. Proc. 15th Genetic and Evolutionary Computation COnference - ACM-GECCO, ACM SIGEVO, Jul 2013, Amsterdam, Netherlands. pp.1517-1524. ⟨hal-01084312⟩