arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2307.04868cs.LG
收藏

利用对齐集解决实例依赖标签噪声问题

Leveraging an Alignment Set in Tackling Instance-Dependent Label Noise

University of Michigan(密歇根大学)

Donna Tjandra, Jenna Wiens

AI总结:

本文提出一种利用锚点对齐集的两阶段学习方法,在实例依赖标签噪声下同时提升判别性能(AUROC)并减轻偏差(AUEOC),在MIMIC-III等任务上优于现有方法。

摘要:

有噪声的训练标签会损害模型性能。大多数旨在解决标签噪声的方法假设标签噪声与输入特征无关。然而在实践中,标签噪声往往是特征依赖或实例依赖的,因此是有偏的(即某些实例比其他实例更可能被错误标注)。例如,在临床护理中,女性患者比男性患者更可能被漏诊心血管疾病。忽略这种依赖性的方法会产生判别性能较差的模型,并且在许多医疗保健场景中,可能加剧健康差异问题。鉴于这些局限性,我们提出了一种两阶段方法,用于在存在实例依赖标签噪声的情况下进行学习。我们的方法利用锚点(anchor points),即一小部分我们知道观测标签和真实标签的数据子集。在多个任务上,我们的方法在判别性能(AUROC)上相对于最先进方法取得了一致改进,同时减轻了偏差(均等化几率曲线下面积,AUEOC)。例如,在MIMIC-III数据集上预测急性呼吸衰竭发作时,我们的方法实现了0.84(标准差[SD] 0.01)的调和平均值(AUROC和AUEOC),而次优基线的调和平均值为0.81(SD 0.01)。总体而言,与现有方法相比,我们的方法在存在实例依赖标签噪声的情况下提高了准确性,同时减轻了潜在偏差。

英文摘要:

Noisy training labels can hurt model performance. Most approaches that aim to address label noise assume label noise is independent from the input features. In practice, however, label noise is often feature or \textit{instance-dependent}, and therefore biased (i.e., some instances are more likely to be mislabeled than others). E.g., in clinical care, female patients are more likely to be under-diagnosed for cardiovascular disease compared to male patients. Approaches that ignore this dependence can produce models with poor discriminative performance, and in many healthcare settings, can exacerbate issues around health disparities. In light of these limitations, we propose a two-stage approach to learn in the presence instance-dependent label noise. Our approach utilizes \textit{\anchor points}, a small subset of data for which we know the observed and ground truth labels. On several tasks, our approach leads to consistent improvements over the state-of-the-art in discriminative performance (AUROC) while mitigating bias (area under the equalized odds curve, AUEOC). For example, when predicting acute respiratory failure onset on the MIMIC-III dataset, our approach achieves a harmonic mean (AUROC and AUEOC) of 0.84 (SD [standard deviation] 0.01) while that of the next best baseline is 0.81 (SD 0.01). Overall, our approach improves accuracy while mitigating potential bias compared to existing approaches in the presence of instance-dependent label noise.

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。

↑