arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2203.10883cs.LG
收藏

极端强盗的有效算法

Efficient Algorithms for Extreme Bandits

CNRS(法国国家科学研究中心);Univ. Lille(里尔大学);Inria(法国国家信息与自动化研究所);Centrale Lille(里尔中央理工学院);ENS(巴黎高等师范学院);Université PSL(巴黎文理研究大学)

Dorian Baudry, Yoan Russac, Emilie Kaufmann

摘要:

在本文中,我们致力于极端强盗问题,一个变种的多武装强盗,其中学习者寻求收集最大可能的奖励。我们首先研究了独立同分布随机变量的最大值在奖励分布尾部的集中性。这种分析促使引入最大分位数(QoMax)。QoMax的属性足以建立一个探索然后提交(ETC)策略,QoMax-ETC,实现强大的渐近保证,尽管它的简单性。然后,我们提出并分析了一种更具自适应性的随时算法QoMax-SDA,它将QoMax与Baudry等人最近引入的子采样方法相结合。这两种算法在两个方面比现有的方法更有效:(1)它们导致更好的经验性能;(2)它们享有显着减少的内存和时间复杂度。

英文摘要:

In this paper, we contribute to the Extreme Bandit problem, a variant of Multi-Armed Bandits in which the learner seeks to collect the largest possible reward. We first study the concentration of the maximum of i.i.d random variables under mild assumptions on the tail of the rewards distributions. This analysis motivates the introduction of Quantile of Maxima (QoMax). The properties of QoMax are sufficient to build an Explore-Then-Commit (ETC) strategy, QoMax-ETC, achieving strong asymptotic guarantees despite its simplicity. We then propose and analyze a more adaptive, anytime algorithm, QoMax-SDA, which combines QoMax with a subsampling method recently introduced by Baudry et al. (2021). Both algorithms are more efficient than existing approaches in two aspects (1) they lead to better empirical performance (2) they enjoy a significant reduction of the memory and time complexities.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑