There has been a recent surge of interest in nonparametric bandit algorithms based on subsampling. One drawback however of these approaches is the additional complexity required by random subsampling and the storage of the full history of rewards. Our first contribution is to show that a simple deterministic subsampling rule, proposed in the recent work of Baudry et al. (2020) under the name of ''last-block subsampling'', is asymptotically optimal in one-parameter exponential families. In addition, we prove that these guarantees also hold when limiting the algorithm memory to a polylogarithmic function of the time horizon. These findings open up new perspectives, in particular for non-stationary scenarios in which the arm distributions evolve over time. We propose a variant of the algorithm in which only the most recent observations are used for subsampling, achieving optimal regret guarantees under the assumption of a known number of abrupt changes. Extensive numerical simulations highlight the merits of this approach, particularly when the changes are not only affecting the means of the rewards.
翻译:最近出现了对基于子抽样的非参数性土匪算法的兴趣激增。 但是,这些方法的一个缺点是随机的子抽样和储存整个奖赏历史所需的额外复杂性。我们的第一个贡献是表明,在Baudry等人(2020年)最近的工作中,以“最后一块次抽样”的名义提出的一个简单的确定性亚抽样规则,在单参数指数家庭中是微不足道的。此外,我们证明,这些保证在将算法记忆限制在时界的多元函数时,也具有一定的优点。这些结论打开了新的视角,特别是对非静止情景而言,在非静止情景中,手臂分布会随着时间的变化而变化。我们提出了一个算法的变式,其中仅使用最近的意见进行次抽样,在已知数突然变化的假设下实现最佳的遗憾保证。广泛的数字模拟突出了这一方法的优点,特别是当这些变化不仅影响奖赏手段时。