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 2301.12386cs.LG
收藏

结合分布外检测的选择性分类的插件估计器

Plugin estimators for selective classification with out-of-distribution detection

Google Research(谷歌研究院)

Harikrishna Narasimhan, Aditya Krishna Menon, Wittawat Jitkrittum, Sanjiv Kumar

AI总结:

针对结合分布外检测的选择性分类(SCOD)理论基础薄弱、现有方法多为启发式的问题,提出有理论依据的新型插件估计器,推广现有方法并取得有竞争力的性能。

摘要:

现实世界中的分类器可受益于在低置信度样本上放弃预测的选项,这种弃权(不执行)在接近已学习决策边界的样本,或是相对于训练样本的离群样本上尤为有用。这些场景在选择性分类(SC)和分布外(OOD)检测文献中已有广泛但相互独立的研究。近期关于结合分布外检测的选择性分类(SCOD)的研究主张对这些问题进行统一研究,但该问题的正式理论基础仍处于初期阶段,现有技术本质上多为启发式方法。本文提出了新的SCOD插件估计器,它们具备理论依据、效果优异,且推广了SC和OOD检测文献中的现有方法。在分析过程中,我们正式阐明了直接套用现有SC和OOD检测基线可能不足以解决SCOD问题。我们通过实验证明,与来自两类文献的基线方法相比,我们的方法能取得具有竞争力的SC和OOD检测性能。

英文摘要:

Real-world classifiers can benefit from the option of abstaining from predicting on samples where they have low confidence. Such abstention is particularly useful on samples which are close to the learned decision boundary, or which are outliers with respect to the training sample. These settings have been the subject of extensive but disjoint study in the selective classification (SC) and out-of-distribution (OOD) detection literature. Recent work on selective classification with OOD detection (SCOD) has argued for the unified study of these problems; however, the formal underpinnings of this problem are still nascent, and existing techniques are heuristic in nature. In this paper, we propose new plugin estimators for SCOD that are theoretically grounded, effective, and generalise existing approaches from the SC and OOD detection literature. In the course of our analysis, we formally explicate how naïve use of existing SC and OOD detection baselines may be inadequate for SCOD. We empirically demonstrate that our approaches yields competitive SC and OOD detection performance compared to baselines from both literatures.

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。

↑