arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2106.04700cs.LGstat.ML
收藏

无标度对抗性多武装土匪

Scale Free Adversarial Multi Armed Bandits

Columbia University(哥伦比亚大学)

Sudeep Raja Putta, Shipra Agrawal

摘要:

我们考虑无标度对抗性多臂强盗(MAB)问题。在游戏开始时,玩家只知道武器$n$的数量。它不知道对手选择的损失的规模和大小,也不知道$T$的回合数。在每一轮中,它看到关于损失向量$l_1,\dots, l_T \in \mathbb{R}^n$的强盗反馈。我们的目标是将其遗憾作为$n$的函数和$l_1,\dots, l_T$的规范进行绑定。我们设计了一个强盗跟随正则化领导者(FTRL)算法,它使用自适应学习率,并给出两个不同的遗憾界,基于所使用的探索参数。对于非自适应探索,我们的算法有一个遗憾$\tilde{\mathcal{O}}(\sqrt{nL_2} + L_\infty\sqrt{nT})$和自适应探索,它有一个遗憾$\tilde{\mathcal{O}}(\sqrt{nL_2} + L_\infty\sqrt{nL_1})$。这里,$L_\infty = \sup_t \| l_t\|_\infty$、$L_2 = \sum_{t=1}^T \|l_t\|_2^2$、$L_1 = \sum_{t=1}^T \|l_t\|_1$和$\tilde{\mathcal{O}}$符号抑制对数因子。这些是适应损失的$\|\cdot\|_2$、$\|\cdot\|_1$范数的第一个MAB界限。第二个界限是第一个数据依赖的无标度MAB界限,因为$T$没有直接出现在遗憾中。我们还开发了一种新的技术,以获得丰富的一类局部范数下界Bregman分歧。这种技术在我们的分析中起着至关重要的作用,控制后悔时,使用重要性加权估计的无限损失。这项技术可能是独立的利益。

英文摘要:

We consider the Scale-Free Adversarial Multi Armed Bandits(MAB) problem. At the beginning of the game, the player only knows the number of arms $n$. It does not know the scale and magnitude of the losses chosen by the adversary or the number of rounds $T$. In each round, it sees bandit feedback about the loss vectors $l_1,\dots, l_T \in \mathbb{R}^n$. The goal is to bound its regret as a function of $n$ and norms of $l_1,\dots, l_T$. We design a bandit Follow The Regularized Leader (FTRL) algorithm, that uses an adaptive learning rate and give two different regret bounds, based on the exploration parameter used. With non-adaptive exploration, our algorithm has a regret of $\tilde{\mathcal{O}}(\sqrt{nL_2} + L_\infty\sqrt{nT})$ and with adaptive exploration, it has a regret of $\tilde{\mathcal{O}}(\sqrt{nL_2} + L_\infty\sqrt{nL_1})$. Here $L_\infty = \sup_t \| l_t\|_\infty$, $L_2 = \sum_{t=1}^T \|l_t\|_2^2$, $L_1 = \sum_{t=1}^T \|l_t\|_1$ and the $\tilde{\mathcal{O}}$ notation suppress logarithmic factors. These are the first MAB bounds that adapt to the $\|\cdot\|_2$, $\|\cdot\|_1$ norms of the losses. The second bound is the first data-dependent scale-free MAB bound as $T$ does not directly appear in the regret. We also develop a new technique for obtaining a rich class of local-norm lower-bounds for Bregman Divergences. This technique plays a crucial role in our analysis for controlling the regret when using importance weighted estimators of unbounded losses. This technique could be of independent interest.

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。

↑