BriefGPT.xyz
Feb, 2024
贝叶斯框架下的固定置信度最佳臂识别
Fixed Confidence Best Arm Identification in the Bayesian Setting
HTML
PDF
Kyoungseok Jang, Junpei Komiyama, Kazutoshi Yamazaki
TL;DR
在贝叶斯设置下，我们研究了固定置信度最佳臂识别问题。我们证明了传统的FC-BAI算法在贝叶斯设置下会导致任意次优的性能，并且介绍了一种连续淘汰的变体，其性能与下界匹配，仅有一个对数因子的差距。模拟实验验证了理论结果。
Abstract
We consider the
fixed-confidence best arm identification
(FC-BAI) problem in the
bayesian setting
. This problem aims to find the arm of the largest mean with a fixed confidence level when the bandit model has bee
→