Learning from Noisy Complementary Labels with Robust Loss Functions
Learning from Noisy Complementary Labels with Robust Loss Functions
复制标题
DOI：
10.1587/transinf.2021edp7035
复制
发表时间：
2021-02
期刊：
影响因子：
--
通讯作者：
Hiroki Ishiguro;Takashi Ishida;Masashi Sugiyama
中科院分区：
文献类型：
--
作者：
Hiroki Ishiguro;Takashi Ishida;Masashi Sugiyama
摘要已经证明，大规模的标记数据集有助于机器学习的成功。然而，在实践中，收集标记数据的成本往往非常高，而且容易出错。为了解决这个问题，以前的研究已经考虑使用补充标签，它指定fi是一个实例不属于的类，并且比普通标签更容易收集。然而，互补标签也可能容易出错，因此缓解标签噪声的干扰是使互补标签学习在实践中更有用的一个重要挑战。在这篇文章中，我们给出了损失函数的条件，使得学习算法不是受互补标签中的噪声影响的ff。在含有噪声互补标签的基准数据集上的实验表明，满足条件的损失函数可以显著提高fi的Classifi阳离子性能。
SUMMARY It has been demonstrated that large-scale labeled datasets facilitate the success of machine learning. However, collecting labeled data is often very costly and error-prone in practice. To cope with this problem, previous studies have considered the use of a complementary label, which specifies a class that an instance does not belong to and can be collected more easily than ordinary labels. However, complementary labels could also be error-prone and thus mitigating the influence of label noise is an important challenge to make complementary-label learning more useful in practice. In this paper, we derive conditions for the loss function such that the learning algorithm is not a ff ected by noise in complementary labels. Experiments on benchmark datasets with noisy complementary labels demonstrate that the loss functions that satisfy our conditions significantly improve the classification performance.