Multi-armed Bandit, Dynamic Environments and Meta-Bandits

Cédric Hartland (1, 2) , Sylvain Gelly (1, 2) , Nicolas Baskiotis (1, 2) , Olivier Teytaud (1, 2) , Michèle Sebag (1, 2)

1 LRI - Laboratoire de Recherche en Informatique
2 TANC - Algorithmic number theory for cryptology

This paper presents the Adapt-EvE algorithm, extending the UCBT online learning algorithm (Auer et al. 2002) to abruptly changing environments. Adapt-EvE features an adaptive change-point detection test based on Page-Hinkley statistics, and two alternative xtra-exploration procedures respectively based on smooth-restart and Meta-Bandits.

Mots clés: multi-armed bandit; statistical learning; ucb

Domaines: Apprentissage [cs.LG]; Intelligence artificielle [cs.AI]

Cédric Hartland, Sylvain Gelly, Nicolas Baskiotis, Olivier Teytaud, Michèle Sebag. Multi-armed Bandit, Dynamic Environments and Meta-Bandits. 2006. ⟨hal-00113668⟩