arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.08862cs.LGcs.DSstat.ML
收藏

近优交换遗憾最小化用于凸损失

Near-optimal Swap Regret Minimization for Convex Losses

Northeastern University(东北大学);Google(谷歌);Microsoft Research(微软研究院)

Lunjia Hu, Jon Schneider, Yifan Wu

AI总结:

本文提出了一种高效的在线算法,实现凸损失下的近优交换遗憾最小化,并首次为中位数校准提供了O(√T)的校准误差保证。

摘要:

我们提出了一种随机在线算法,能够保证在任何T个自适应选择的Lipschitz凸损失序列上,预期交换遗憾达到近最优的~O(√T)。这改进了之前的最佳界~O(T^{2/3}),并回答了Fishelson等人[2025b]提出的一个开放问题。此外,我们的算法是高效的:它在poly(T)时间内运行。我们开发的一项关键技术思想是将单位区间离散化为多个粒度的bin,并同时利用所有粒度来做出随机预测,这被称为多尺度分箱,可能具有独立的兴趣。我们结果的直接推论是为一般可 elicitable 属性最小化校准误差的高效在线算法。这一结果不需要先前工作中所需的识别函数的Lipschitz性假设,使其适用于中位数校准,我们为此达到了首个~O(√T)校准误差保证。

英文摘要:

We give a randomized online algorithm that guarantees near-optimal $\widetilde O(\sqrt T)$ expected swap regret against any sequence of $T$ adaptively chosen Lipschitz convex losses on the unit interval. This improves the previous best bound of $\widetilde O(T^{2/3})$ and answers an open question of Fishelson et al. [2025b]. In addition, our algorithm is efficient: it runs in $\mathsf{poly}(T)$ time. A key technical idea we develop to obtain this result is to discretize the unit interval into bins at multiple scales of granularity and simultaneously use all scales to make randomized predictions, which we call multi-scale binning and may be of independent interest. A direct corollary of our result is an efficient online algorithm for minimizing the calibration error for general elicitable properties. This result does not require the Lipschitzness assumption of the identification function needed in prior work, making it applicable to median calibration, for which we achieve the first $\widetilde O(\sqrt T)$ calibration error guarantee.

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。

↑