Noisy Feature Mixup

TL;DR

Noisy Feature Mixup (NFM) is a data augmentation technique that combines interpolation and noise injection in input and feature space, improving model robustness and decision boundary smoothing.

Contribution

This paper introduces NFM, a novel augmentation method that generalizes mixup techniques and enhances robustness and decision boundary smoothing, supported by theoretical analysis and empirical validation.

Findings

NFM improves robustness across vision benchmarks.

NFM achieves better trade-offs between accuracy and robustness.

NFM outperforms mixup and manifold mixup in experiments.

Abstract

We introduce Noisy Feature Mixup (NFM), an inexpensive yet effective method for data augmentation that combines the best of interpolation based training and noise injection schemes. Rather than training with convex combinations of pairs of examples and their labels, we use noise-perturbed convex combinations of pairs of data points in both input and feature space. This method includes mixup and manifold mixup as special cases, but it has additional advantages, including better smoothing of decision boundaries and enabling improved model robustness. We provide theory to understand this as well as the implicit regularization effects of NFM. Our theory is supported by empirical results, demonstrating the advantage of NFM, as compared to mixup and manifold mixup. We show that residual networks and vision transformers trained with NFM have favorable trade-offs between predictive accuracy on clean data and robustness with respect to various types of data perturbation across a range of computer vision benchmark datasets.