Fast active learning for pure exploration in reinforcement learning - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Rapport (Rapport De Recherche) Année : 2020

Fast active learning for pure exploration in reinforcement learning

Pierre Ménard
  • Fonction : Auteur
  • PersonId : 1022182
Emilie Kaufmann
Edouard Leurent
  • Fonction : Auteur
  • PersonId : 1038135
Michal Valko

Résumé

Realistic environments often provide agents with very limited feedback. When the environment is initially unknown, the feedback, in the beginning, can be completely absent, and the agents may first choose to devote all their effort on exploring efficiently. The exploration remains a challenge while it has been addressed with many hand-tuned heuristics with different levels of generality on one side, and a few theoretically backed exploration strategies on the other. Many of them are incarnated by intrinsic motivation and in particular explorations bonuses. A common rule of thumb for exploration bonuses is to use 1/ √ n bonus that is added to the empirical estimates of the reward, where n is a number of times this particular state (or a state-action pair) was visited. We show that, surprisingly, for a pure-exploration objective of reward-free exploration, bonuses that scale with 1/n bring faster learning rates, improving the known upper bounds with respect to the dependence on the horizon H. Furthermore, we show that with an improved analysis of the stopping time, we can improve by a factor H the sample complexity in the best-policy identification setting, which is another pure-exploration objective, where the environment provides rewards but the agent is not penalized for its behavior during the exploration phase.
Fichier principal
Vignette du fichier
ActiveExplorationMDP.pdf (812.93 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02906985 , version 1 (26-07-2020)
hal-02906985 , version 2 (27-07-2020)
hal-02906985 , version 3 (16-07-2021)

Identifiants

  • HAL Id : hal-02906985 , version 2

Citer

Pierre Ménard, Omar Darwiche Domingues, Anders Jonsson, Emilie Kaufmann, Edouard Leurent, et al.. Fast active learning for pure exploration in reinforcement learning. [Research Report] DeepMind. 2020. ⟨hal-02906985v2⟩

Collections

LARA
281 Consultations
244 Téléchargements

Partager

Gmail Facebook X LinkedIn More