Skip to Main content Skip to Navigation
Conference papers

Change Point Detection and Meta-Bandits for Online Learning in Dynamic Environments

Cédric Hartland 1, 2 Nicolas Baskiotis 1, 2 Sylvain Gelly 1, 2 Michèle Sebag 1, 2 Olivier Teytaud 1, 2
1 TANC - Algorithmic number theory for cryptology
Inria Saclay - Ile de France, LIX - Laboratoire d'informatique de l'École polytechnique [Palaiseau]
Abstract : Motivated by realtime website optimization, this paper is about online learning in abruptly changing environments. Two extensions of the UCBT algorithm are combined in order to handle dynamic multi-armed bandits, and specifically to cope with fast variations in the rewards. Firstly, a change point detection test based on Page-Hinkley statistics is used to overcome the limitations due to the UCBT inertia. Secondly, a controlled forgetting strategy dubbed Meta-Bandit is proposed to take care of the Exploration vs Exploitation trade-off when the PH test is triggered. Extensive empirical validation shows significant improvements compared to the baseline algorithms. The paper also investigates the sensitivity of the proposed algorithm with respect to the number of available options.
Complete list of metadatas

Cited literature [13 references]  Display  Hide  Download

https://hal.inria.fr/inria-00164033
Contributor : Cédric Hartland <>
Submitted on : Monday, November 5, 2007 - 10:18:05 AM
Last modification on : Tuesday, June 23, 2020 - 11:34:26 AM
Document(s) archivé(s) le : Monday, September 24, 2012 - 11:15:50 AM

File

cap07.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : inria-00164033, version 1

Collections

Citation

Cédric Hartland, Nicolas Baskiotis, Sylvain Gelly, Michèle Sebag, Olivier Teytaud. Change Point Detection and Meta-Bandits for Online Learning in Dynamic Environments. CAp 2007 : 9è Conférence francophone sur l'apprentissage automatique, Jul 2007, Grenoble, France. pp.237-250. ⟨inria-00164033⟩

Share

Metrics

Record views

885

Files downloads

1380