How Does Mixup Help With Robustness and Generalization?

TL;DR

This paper provides a theoretical framework explaining how Mixup improves model robustness against adversarial attacks and enhances generalization by reducing overfitting through data-adaptive regularization.

Contribution

It offers the first theoretical analysis linking Mixup to robustness and generalization, demonstrating its effects on adversarial loss and overfitting reduction.

Findings

Mixup minimizes an upper bound of adversarial loss, enhancing robustness.

Mixup acts as a data-adaptive regularizer, reducing overfitting.

Theoretical insights explain Mixup's effectiveness in improving model performance.

Abstract

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.