Discover and Cure: Concept-aware Mitigation of Spurious Correlation-猫眼课题宝

DOI：

10.48550/arxiv.2305.00650

发表时间：

2023-05

期刊：

ArXiv

影响因子：

通讯作者：

中科院分区：

其他

文献类型：

作者：

Shirley Wu;Mert Yuksekgonul;Linjun Zhang;James Y. Zou

深度神经网络通常依赖于虚假的相关性来进行预测，这阻碍了训练环境之外的泛化。例如，将猫与床背景相关联的模型可能无法预测猫在没有床的其他环境中的存在。减少虚假的相关性对于构建可信的模型至关重要。然而，现有的工程缺乏透明度，无法让人深入了解缓解过程。在这项工作中，我们提出了一个可解释的框架，发现和治愈（DISC），来解决这个问题。对于人类可解释的概念，DISC迭代地1）在不同环境中发现不稳定的概念作为虚假属性，然后2）使用发现的概念干预训练数据以减少虚假相关性。在系统的实验中，DISC提供了优于现有方法的泛化能力和可解释性。具体来说，它在对象识别任务和皮肤病变分类任务上的性能分别比最先进的方法高出7.5%和9.6%。此外，我们还提供理论分析和保证，以了解DISC训练的模型的好处。代码和数据可在https://github.com/Wuyxin/DISC上获得。

Deep neural networks often rely on spurious correlations to make predictions, which hinders generalization beyond training environments. For instance, models that associate cats with bed backgrounds can fail to predict the existence of cats in other environments without beds. Mitigating spurious correlations is crucial in building trustworthy models. However, the existing works lack transparency to offer insights into the mitigation process. In this work, we propose an interpretable framework, Discover and Cure (DISC), to tackle the issue. With human-interpretable concepts, DISC iteratively 1) discovers unstable concepts across different environments as spurious attributes, then 2) intervenes on the training data using the discovered concepts to reduce spurious correlation. Across systematic experiments, DISC provides superior generalization ability and interpretability than the existing approaches. Specifically, it outperforms the state-of-the-art methods on an object recognition task and a skin-lesion classification task by 7.5% and 9.6%, respectively. Additionally, we offer theoretical analysis and guarantees to understand the benefits of models trained by DISC. Code and data are available at https://github.com/Wuyxin/DISC.

Discover and Cure: Concept-aware Mitigation of Spurious Correlation

文献摘要

相似文献