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 2103.11489cs.LGstat.ML
收藏

基于UCB的多项Logistic回归Bandits算法

UCB-based Algorithms for Multinomial Logistic Regression Bandits

University of California, Los Angeles(加利福尼亚大学洛杉矶分校);University of British Columbia(不列颠哥伦比亚大学)

Sanae Amani, Christos Thrampoulidis

摘要:

在广义线性强盗的丰富家族中,也许研究得最好的是在二元奖励问题中使用的逻辑强盗:例如,当学习者/代理试图最大化可以选择两种可能结果之一的用户的利润时(例如,“click”vs“no-click”)。尽管最近取得了显著的进展并且改进了用于物流强盗的算法,但是现有的工作没有解决用户可以选择的结果的数量大于两个的实际情况(例如,“click”、“show me later”、“never show again”、“no click”)。在本文中,我们研究这样的扩展。我们使用多项logit(MNL)来模拟$K+1\geq 2$可能结果中的每一个的概率(+1代表“不点击”结果):我们假设对于学习者的动作$\mathbf{x}_t$,用户选择$K+1\geq 2$结果中的一个,比如结果$i$,具有多项logit(MNL)概率模型,具有相应的未知参数$\bar{\boldsymbolθ}_{\ast i}$。每个结果$i$还与收入参数$ρ_i$相关联,目标是最大化预期收入。对于这个问题,我们提出了MNL-UCB,上置信界(UCB)为基础的算法,实现遗憾$\tilde{\mathcal{O}}(dK\sqrt{T})$与小依赖问题相关的常数,否则可以任意大,导致松散的遗憾界限。我们提出的数值模拟,证实了我们的理论结果。

英文摘要:

Out of the rich family of generalized linear bandits, perhaps the most well studied ones are logisitc bandits that are used in problems with binary rewards: for instance, when the learner/agent tries to maximize the profit over a user that can select one of two possible outcomes (e.g., `click' vs `no-click'). Despite remarkable recent progress and improved algorithms for logistic bandits, existing works do not address practical situations where the number of outcomes that can be selected by the user is larger than two (e.g., `click', `show me later', `never show again', `no click'). In this paper, we study such an extension. We use multinomial logit (MNL) to model the probability of each one of $K+1\geq 2$ possible outcomes (+1 stands for the `not click' outcome): we assume that for a learner's action $\mathbf{x}_t$, the user selects one of $K+1\geq 2$ outcomes, say outcome $i$, with a multinomial logit (MNL) probabilistic model with corresponding unknown parameter $\bar{\boldsymbolθ}_{\ast i}$. Each outcome $i$ is also associated with a revenue parameter $ρ_i$ and the goal is to maximize the expected revenue. For this problem, we present MNL-UCB, an upper confidence bound (UCB)-based algorithm, that achieves regret $\tilde{\mathcal{O}}(dK\sqrt{T})$ with small dependency on problem-dependent constants that can otherwise be arbitrarily large and lead to loose regret bounds. We present numerical simulations that corroborate our theoretical results.

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。

↑