Noisy Feature Mixup

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.

34 pages

References in corpus (14)

• Explaining and Harnessing Adversarial Examples
• Theoretically Principled Trade-off between Robustness and Accuracy
• AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty
• Escaping the Big Data Paradigm with Compact Transformers
• Robustness of classifiers: from adversarial to random noise
• Fixing Data Augmentation to Improve Adversarial Robustness
• Puzzle Mix: Exploiting Saliency and Local Statistics for Optimal Mixup
• Theoretical evidence for adversarial robustness through randomization
• MaxUp: A Simple Way to Improve Generalization of Neural Network Training
• How Does Mixup Help With Robustness and Generalization?
• When and How Mixup Improves Calibration
• Noisy Recurrent Neural Networks
• A unified view on differential privacy and robustness to adversarial examples
• k-Mixup Regularization for Deep Learning via Optimal Transport