New parameterized algorithm families for improving multi-armed bandits achieve optimal dependence on the number of arms under sufficient concavity and provide best-arm identification on well-behaved instances while reverting to worst-case guarantees otherwise.
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
UNVERDICTED 1representative citing papers
citing papers explorer
-
Algorithm Design and Stronger Guarantees for the Improving Multi-Armed Bandits Problem
New parameterized algorithm families for improving multi-armed bandits achieve optimal dependence on the number of arms under sufficient concavity and provide best-arm identification on well-behaved instances while reverting to worst-case guarantees otherwise.