Logo du site
  • English
  • Français
  • Se connecter
Logo du site
  • English
  • Français
  • Se connecter
  1. Accueil
  2. Université de Neuchâtel
  3. Publications
  4. Randomised Bayesian Least-Squares Policy Iteration
 
  • Details
Options
Vignette d'image

Randomised Bayesian Least-Squares Policy Iteration

Auteur(s)
Nikolaos Tziortziotis
Dimitrakakis, Christos 
Institut d'informatique 
Michalis Vazirgiannis
Date de parution
2019-04-06T21:50:24Z
In
Computing Research Repository (CoRR)
Vol.
1904.03535
Mots-clés
  • cs.LG
  • cs.AI
  • stat.ML
  • cs.LG

  • cs.AI

  • stat.ML

Résumé
We introduce Bayesian least-squares policy iteration (BLSPI), an off-policy, model-free, policy iteration algorithm that uses the Bayesian least-squares temporal-difference (BLSTD) learning algorithm to evaluate policies. An online variant of BLSPI has been also proposed, called randomised BLSPI (RBLSPI), that improves its policy based on an incomplete policy evaluation step. In online setting, the exploration-exploitation dilemma should be addressed as we try to discover the optimal policy by using samples collected by ourselves. RBLSPI exploits the advantage of BLSTD to quantify our uncertainty about the value function. Inspired by Thompson sampling, RBLSPI first samples a value function from a posterior distribution over value functions, and then selects actions based on the sampled value function. The effectiveness and the exploration abilities of RBLSPI are demonstrated experimentally in several environments.
Identifiants
https://libra.unine.ch/handle/123456789/30985
_
10.48550/arXiv.1904.03535
_
1904.03535v1
Type de publication
journal article
Dossier(s) à télécharger
 main article: 1904.03535.pdf (7.63 MB)
google-scholar
Présentation du portailGuide d'utilisationStratégie Open AccessDirective Open Access La recherche à l'UniNE Open Access ORCIDNouveautés

Service information scientifique & bibliothèques
Rue Emile-Argand 11
2000 Neuchâtel
contact.libra@unine.ch

Propulsé par DSpace, DSpace-CRIS & 4Science | v2022.02.00