On Non-Random Missing Labels in Semi-Supervised Learning
On Non-Random Missing Labels in Semi-Supervised Learning
复制标题
关于半监督学习中的非随机缺失标签
DOI：
--
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
Hanwang Zhang
中科院分区：
文献类型：
--
作者：
Xinting Hu;Yulei Niu;C. Miao;Xiansheng Hua;Hanwang Zhang
半监督学习（SSL）从根本上来说是一个缺失标签问题，其中标签非随机缺失（MNAR）问题比广泛采用但幼稚的完全随机缺失假设（其中标记和未标记数据共享相同的类分布）更现实和更具挑战性。与现有的 SSL 解决方案忽视“类”在导致非随机性中的作用（例如，用户更有可能标记流行类）不同，我们明确地将“类”合并到 SSL 中。我们的方法有三部分：1）我们提出了类别感知倾向（CAP），它利用未标记的数据来使用有偏差的标记数据来训练改进的分类器。 2）为了鼓励稀有类训练，其模型召回率低但精度高，会丢弃太多伪标签数据，我们提出了类感知插补（CAI），它动态降低（或增加）稀有（或频繁）类的伪标签分配阈值。 3) 总体而言，我们将 CAP 和 CAI 集成到类感知双鲁棒 (CADR) 估计器中，用于训练无偏 SSL 模型。在各种 MNAR 设置和消融下，我们的方法不仅显着优于现有基线，而且还超过了其他标签偏差消除 SSL 方法。请检查我们的代码：https://github.com/JoyHuYY1412/CADR-FixMatch。
Semi-Supervised Learning (SSL) is fundamentally a missing label problem, in which the label Missing Not At Random (MNAR) problem is more realistic and challenging, compared to the widely-adopted yet naive Missing Completely At Random assumption where both labeled and unlabeled data share the same class distribution. Different from existing SSL solutions that overlook the role of"class"in causing the non-randomness, e.g., users are more likely to label popular classes, we explicitly incorporate"class"into SSL. Our method is three-fold: 1) We propose Class-Aware Propensity (CAP) that exploits the unlabeled data to train an improved classifier using the biased labeled data. 2) To encourage rare class training, whose model is low-recall but high-precision that discards too many pseudo-labeled data, we propose Class-Aware Imputation (CAI) that dynamically decreases (or increases) the pseudo-label assignment threshold for rare (or frequent) classes. 3) Overall, we integrate CAP and CAI into a Class-Aware Doubly Robust (CADR) estimator for training an unbiased SSL model. Under various MNAR settings and ablations, our method not only significantly outperforms existing baselines but also surpasses other label bias removal SSL methods. Please check our code at: https://github.com/JoyHuYY1412/CADR-FixMatch.
影响因子：
5
作者：
Funk, Michele Jonsson;Westreich, Daniel;Davidian, Marie
通讯作者：
Davidian, Marie