How Does Mixup Help With Robustness and Generalization?
This work addresses a fundamental gap in explaining a widely used data augmentation technique for machine learning practitioners, though it is incremental as it builds on existing Mixup methods without introducing new paradigms.
The paper tackles the lack of understanding of why Mixup improves model robustness and generalization by providing theoretical analysis, showing that minimizing Mixup loss approximates an upper bound of adversarial loss for robustness and acts as data-adaptive regularization to reduce overfitting for generalization.
Mixup is a popular data augmentation technique based on taking convex combinations of pairs of examples and their labels. This simple technique has been shown to substantially improve both the robustness and the generalization of the trained model. However, it is not well-understood why such improvement occurs. In this paper, we provide theoretical analysis to demonstrate how using Mixup in training helps model robustness and generalization. For robustness, we show that minimizing the Mixup loss corresponds to approximately minimizing an upper bound of the adversarial loss. This explains why models obtained by Mixup training exhibits robustness to several kinds of adversarial attacks such as Fast Gradient Sign Method (FGSM). For generalization, we prove that Mixup augmentation corresponds to a specific type of data-adaptive regularization which reduces overfitting. Our analysis provides new insights and a framework to understand Mixup.