Toward Learning Human-aligned Cross-domain Robust Models by Countering Misaligned Features
Toward Learning Human-aligned Cross-domain Robust Models by Countering Misaligned Features
复制标题
DOI：
--
复制
发表时间：
2021-11
期刊：
影响因子：
--
通讯作者：
Haohan Wang;Zeyi Huang;Hanlin Zhang;Eric P. Xing
中科院分区：
文献类型：
--
作者：
Haohan Wang;Zeyi Huang;Hanlin Zhang;Eric P. Xing
机器学习已经证明了对i.i.d数据的显著预测准确性，但当使用来自另一个分布的数据进行测试时，准确性往往会下降。在本文中，我们的目标是从另一个角度来看待这个问题，假设这种准确性下降背后的原因是模型对特征的依赖，这些特征与数据注释器在这两个数据集上的相似性没有很好地保持一致。我们将这些特征称为未对准特征。我们将传统的泛化误差范围扩展到一个新的，这种设置与知识的错位功能是如何与标签相关联。我们的分析为这个问题提供了一套技术，这些技术自然与鲁棒机器学习文献中的许多以前的方法相关联。我们还比较了这些方法的经验强度，证明了当这些先前的技术结合在一起时的性能，并在https://github.com/OoDBag/WR上实现
Machine learning has demonstrated remarkable prediction accuracy over i.i.d data, but the accuracy often drops when tested with data from another distribution. In this paper, we aim to offer another view of this problem in a perspective assuming the reason behind this accuracy drop is the reliance of models on the features that are not aligned well with how a data annotator considers similar across these two datasets. We refer to these features as misaligned features. We extend the conventional generalization error bound to a new one for this setup with the knowledge of how the misaligned features are associated with the label. Our analysis offers a set of techniques for this problem, and these techniques are naturally linked to many previous methods in robust machine learning literature. We also compared the empirical strength of these methods demonstrated the performance when these previous techniques are combined, with an implementation available at https://github.com/OoDBag/WR