A new algorithm, Sequential-Explore-Exploit, achieves near-optimal non-asymptotic sample complexity for 1-identification in multi-armed bandits, matching lower bounds up to polylog factors in both positive and negative instances.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Near Optimal Non-asymptotic Sample Complexity of 1-Identification
A new algorithm, Sequential-Explore-Exploit, achieves near-optimal non-asymptotic sample complexity for 1-identification in multi-armed bandits, matching lower bounds up to polylog factors in both positive and negative instances.