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 2501.01844cs.LG
收藏

从具有硬标签的模糊数据中学习

Learning from Ambiguous Data with Hard Labels

HKUST(GZ)(香港科技大学(广州));University of British Columbia(不列颠哥伦比亚大学);University of Tübingen(蒂宾根大学);Harbin Institute of Technology(哈尔滨工业大学);Beijing Institute of Mathematical Sciences and Applications(北京数学科学与应用研究院);VecML Inc.(VecML公司)

Zeke Xie, Zheng He, Nan Lu, Lichen Bai, Bao Li, Shuo Yang, Mingming Sun, Ping Li

AI总结:

针对硬标签模糊数据导致模型过度自信的问题,提出 QLL 框架与 CPU 风险估计器,并通过混合生成数据验证其泛化提升。

摘要:

现实世界的数据通常包含内在的模糊性,而常见的单一硬标签标注范式忽视了这一点。使用带有这些硬标签的模糊数据进行标准训练,可能会产生过度自信的模型,从而导致泛化能力较差。在本文中,我们提出了一个名为 Quantized Label Learning(QLL,量化标签学习)的新颖框架来缓解这一问题。首先,我们将 QLL 表述为从带有硬标签的(非常)模糊的数据中学习:理想情况下,每个模糊实例都应关联一个真实软标签分布,用于描述其在每个类别中对应的概率权重;然而,这通常无法获得;在实践中,我们只能观察到每个实例的一个量化标签,即从相应的真实软标签分布中采样(量化)得到的硬标签,这可以看作是对真实软标签的有偏近似。其次,我们提出了一种 Class-wise Positive-Unlabeled(CPU,类别的正例-未标注)风险估计器,使我们能够仅从带有量化标签的模糊数据中训练准确的分类器。第三,为了模拟现实世界中带有量化标签的模糊数据集,我们设计了一种基于混合的模糊数据生成过程用于实证评估。实验表明,我们的 CPU 方法能够显著提升模型的泛化性能,并优于基线方法。

英文摘要:

Real-world data often contains intrinsic ambiguity that the common single-hard-label annotation paradigm ignores. Standard training using ambiguous data with these hard labels may produce overly confident models and thus leading to poor generalization. In this paper, we propose a novel framework called Quantized Label Learning (QLL) to alleviate this issue. First, we formulate QLL as learning from (very) ambiguous data with hard labels: ideally, each ambiguous instance should be associated with a ground-truth soft-label distribution describing its corresponding probabilistic weight in each class, however, this is usually not accessible; in practice, we can only observe a quantized label, i.e., a hard label sampled (quantized) from the corresponding ground-truth soft-label distribution, of each instance, which can be seen as a biased approximation of the ground-truth soft-label. Second, we propose a Class-wise Positive-Unlabeled (CPU) risk estimator that allows us to train accurate classifiers from only ambiguous data with quantized labels. Third, to simulate ambiguous datasets with quantized labels in the real world, we design a mixing-based ambiguous data generation procedure for empirical evaluation. Experiments demonstrate that our CPU method can significantly improve model generalization performance and outperform the baselines.

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。

↑