arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2201.13128cs.DScs.LGstat.ML
收藏

拟阵上的删除鲁棒子模最大化

Deletion Robust Submodular Maximization over Matroids

Google Research(谷歌研究院);Sapienza University of Rome(罗马第一大学)

Paul Dütting, Federico Fusco, Silvio Lattanzi, Ashkan Norouzi-Fard, Morteza Zadimoghaddam

AI总结:

本文研究了经典拟阵约束下的单调子模最大化删除鲁棒问题,提出了空间复杂度依赖于拟阵秩和删除元素数量的常数因子近似算法,并在集中式和流式设置下分别给出了近似比与摘要大小保证,实验验证了其有效性。

摘要:

最大化单调子模函数是机器学习中的一项基本任务。本文研究经典拟阵约束下该问题的删除鲁棒版本。目标是提取数据集的小规模摘要,即使敌手删除了部分元素,该摘要仍包含高价值的独立集。我们提出了常数因子近似算法,其空间复杂度取决于拟阵的秩 $k$ 和被删除元素的数量 $d$。在集中式设置下,我们提出了一种 $(3.582+O(\varepsilon))$-近似算法,摘要大小为 $O(k + \frac{d \log k}{\varepsilon^2})$。在流式设置下,我们提供了一种 $(5.582+O(\varepsilon))$-近似算法,其摘要大小和内存均为 $O(k + \frac{d \log k}{\varepsilon^2})$。我们通过深入的实验分析补充了理论结果,展示了算法在真实数据集上的有效性。

英文摘要:

Maximizing a monotone submodular function is a fundamental task in machine learning. In this paper, we study the deletion robust version of the problem under the classic matroids constraint. Here the goal is to extract a small size summary of the dataset that contains a high value independent set even after an adversary deleted some elements. We present constant-factor approximation algorithms, whose space complexity depends on the rank $k$ of the matroid and the number $d$ of deleted elements. In the centralized setting we present a $(3.582+O(\varepsilon))$-approximation algorithm with summary size $O(k + \frac{d \log k}{\varepsilon^2})$. In the streaming setting we provide a $(5.582+O(\varepsilon))$-approximation algorithm with summary size and memory $O(k + \frac{d \log k}{\varepsilon^2})$. We complement our theoretical results with an in-depth experimental analysis showing the effectiveness of our algorithms on real-world datasets.

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。

↑