Analyse non asymptotique d'un test séquentiel de détection de rupture et application aux bandits non stationnaires

Lilian Besson 1, 2, 3, 4, 5 Emilie Kaufmann 6, 3
Abstract : We study a strategy for online change-point detection based on generalized likelihood ratios (GLR) and that can be expressed with the binary relative entropy. This test is used to detect a change in the mean of a bounded distribution, and we propose a non-asymptotic control of its false alarm probability and detection delay. We then explain how it can be useful for sequential decision making by proposing the GLR-klUCB bandit strategy, which is efficient in piece-wise stationary multi-armed bandit models.
Complete list of metadatas

Cited literature [13 references]  Display  Hide  Download

https://hal.inria.fr/hal-02152243
Contributor : Lilian Besson <>
Submitted on : Tuesday, June 11, 2019 - 11:50:42 AM
Last modification on : Wednesday, July 10, 2019 - 1:07:18 AM

File

GRESTI__2019__Paper__Lilian_Be...
Files produced by the author(s)

Licence


Distributed under a Creative Commons Attribution - NonCommercial - ShareAlike 4.0 International License

Identifiers

  • HAL Id : hal-02152243, version 1

Citation

Lilian Besson, Emilie Kaufmann. Analyse non asymptotique d'un test séquentiel de détection de rupture et application aux bandits non stationnaires. GRETSI 2019 - XXVIIème Colloque francophone de traitement du signal et des images, Aug 2019, Lille, France. ⟨hal-02152243⟩

Share

Metrics

Record views

45

Files downloads

526