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 2206.09375cs.LG
收藏

从含分布外样本的非独立同分布数据中进行灰度学习(Gray Learning)

Gray Learning from Non-IID Data with Out-of-distribution Samples

Macquarie University(麦考瑞大学);Sun Yat-sen University(中山大学)

Zhilin Zhao, Longbing Cao, Chang-Dong Wang

AI总结:

针对含分布外样本的非独立同分布数据中标签不可靠、鲁棒学习困难的问题,提出Gray Learning方法,结合真实与互补标签并自适应调权,经理论证明泛化误差界紧致,实验效果显著优于鲁棒统计类方法。

摘要:

训练数据的完整性即便由专家标注也远无法得到保证,尤其是在同时包含分布内与分布外样本的非独立同分布(non-IID)数据集中更是如此。理想情况下,大多数样本应为分布内样本,而语义上存在偏差的样本会在标注过程中被识别为分布外样本并被排除。然而,专家可能会错误地将这些分布外样本归类为分布内样本,为其分配本质上并不可靠的标签。这种不可靠标签与多样数据类型混合的情况,使得训练鲁棒神经网络的任务极具挑战性。我们观察到,除了对应不可靠真实标签的类别之外,分布内和分布外样本几乎总能被排除在某些类别之外。这为利用可靠的互补标签(即指示样本不属于哪些类别的标签)提供了可能。受这一洞见启发,我们提出了一种名为\textit{Gray Learning}(GL,灰度学习)的新方法,该方法同时利用真实标签和互补标签。关键在于,GL会根据预测置信度水平自适应调整这两种标签类型的损失权重。通过将我们的方法建立在统计学习理论基础之上,我们推导得出了泛化误差界,证明GL即使在非独立同分布设置下也能实现紧致的约束边界。大量实验评估表明,我们的方法性能显著优于基于鲁棒统计学的其他替代方法。

英文摘要:

The integrity of training data, even when annotated by experts, is far from guaranteed, especially for non-IID datasets comprising both in- and out-of-distribution samples. In an ideal scenario, the majority of samples would be in-distribution, while samples that deviate semantically would be identified as out-of-distribution and excluded during the annotation process. However, experts may erroneously classify these out-of-distribution samples as in-distribution, assigning them labels that are inherently unreliable. This mixture of unreliable labels and varied data types makes the task of learning robust neural networks notably challenging. We observe that both in- and out-of-distribution samples can almost invariably be ruled out from belonging to certain classes, aside from those corresponding to unreliable ground-truth labels. This opens the possibility of utilizing reliable complementary labels that indicate the classes to which a sample does not belong. Guided by this insight, we introduce a novel approach, termed \textit{Gray Learning} (GL), which leverages both ground-truth and complementary labels. Crucially, GL adaptively adjusts the loss weights for these two label types based on prediction confidence levels. By grounding our approach in statistical learning theory, we derive bounds for the generalization error, demonstrating that GL achieves tight constraints even in non-IID settings. Extensive experimental evaluations reveal that our method significantly outperforms alternative approaches grounded in robust statistics.

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。

↑