Early stopping rules for linear contextual bandits are derived from regret upper bounds and from estimated estimator variances, with a proposed post-stopping conditional inference procedure.
Statistical inference for online decision making: In a contextual bandit setting
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
math.ST 1years
2025 1verdicts
REJECT 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Early Stopping in Contextual Bandits and Inferences
Early stopping rules for linear contextual bandits are derived from regret upper bounds and from estimated estimator variances, with a proposed post-stopping conditional inference procedure.