hal-00595695, version 3
Robust approachability and regret minimization in games with partial monitoring
(23/01/2012)
Résumé : Approachability has become a standard tool in analyzing earning algorithms in the adversarial online learning setup. We develop a variant of approachability for games where there is ambiguity in the obtained reward that belongs to a set, rather than being a single vector. Using this variant we tackle the problem of approachability in games with partial monitoring and develop simple and efficient algorithms (i.e., with constant per-step complexity) for this setup. We finally consider external regret and internal regret in repeated games with partial monitoring and derive regret-minimizing strategies based on approachability theory.
- 1 :
- Israel Institute of Technology
- 2 :
- CNRS : UMR8536 – École normale supérieure de Cachan - ENS Cachan
- 3 :
- CNRS : UMR8553 – Ecole normale supérieure de Paris - ENS Paris
- 4 :
- GROUPE HEC – CNRS : UMR2959
- 5 :
- Ecole normale supérieure de Paris - ENS Paris – INRIA
- Domaine : Mathématiques/Statistiques
Statistiques/Théorie
Informatique/Apprentissage - Versions disponibles : v1 (25-05-2011) v2 (30-08-2011) v3 (15-02-2012)
- hal-00595695, version 3
- http://hal.archives-ouvertes.fr/hal-00595695
- oai:hal.archives-ouvertes.fr:hal-00595695
- Contributeur :
- Soumis le : Mercredi 15 Février 2012, 15:19:50
- Dernière modification le : Mercredi 15 Février 2012, 15:38:49



Documents associés

Exporter