Effects of Mixed Sample Data Augmentation on Interpretability of Neural Networks
Soyoun Won Sung-Ho Bae Seong Tae Kim
Abstract
Mixed sample data augmentation strategies are actively used when training deep neural networks (DNNs). Recent studies suggest that they are effective at various tasks. However, the impact of mixed sample data augmentation on model interpretability has not been widely studied. In this paper, we explore the relationship between model interpretability and mixed sample data augmentation, specifically in terms of feature attribution maps. To this end, we introduce a new metric that allows a comparison of model interpretability while minimizing the impact of occlusion robustness of the model. Experimental results show that several mixed sample data augmentation decreases the interpretability of the model and label mixing during data augmentation plays a significant role in this effect. This new finding suggests it is important to carefully adopt the mixed sample data augmentation method, particularly in applications where attribution map-based interpretability is important.
中文速览
混合样本数据增强（mixed sample data augmentation）是训练深度神经网络时常用的技巧，能提升模型泛化性能，但它对模型可解释性的影响却几乎无人研究。研究者通过特征归因图（feature attribution map）这一视角，系统考察了Mixup、CutMix、SaliencyMix等多种混合增强策略对模型可解释性的影响，并针对现有评估指标存在的缺陷——传统特征遮蔽（feature ablation）评分会被模型的遮挡鲁棒性所干扰，导致比较不公平——提出了一种新指标"跨模型删除"（Inter-Model Deletion），专门用于在排除遮挡鲁棒性偏差的前提下公平比较不同模型的可解释性。实验在CIFAR-10和ImageNet上的结果表明，多种混合增强策略训练出的模型所产生的归因热图质量明显下降，且标签混合（label mixing）是导致可解释性降低的关键因素。这一发现提醒研究者和从业者：在医疗影像、自动驾驶等对模型解释性有严格要求的场景中，混合数据增强方法的选用需要格外谨慎。
原文 arXiv:2303.14608;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2303.14608v2