On the Generalization Effects of Linear Transformations in Data Augmentation
This provides theoretical insights into data augmentation for machine learning practitioners, though it is incremental as it builds on existing augmentation methods.
The paper tackles the problem of understanding why data augmentation works by analyzing linear transformations in over-parametrized linear regression, showing they improve estimation through label preservation and regularization, and validates this with a new augmentation scheme that outperforms random sampling by 1.24% on CIFAR-100 and matches state-of-the-art methods on multiple datasets.
Data augmentation is a powerful technique to improve performance in applications such as image and text classification tasks. Yet, there is little rigorous understanding of why and how various augmentations work. In this work, we consider a family of linear transformations and study their effects on the ridge estimator in an over-parametrized linear regression setting. First, we show that transformations that preserve the labels of the data can improve estimation by enlarging the span of the training data. Second, we show that transformations that mix data can improve estimation by playing a regularization effect. Finally, we validate our theoretical insights on MNIST. Based on the insights, we propose an augmentation scheme that searches over the space of transformations by how uncertain the model is about the transformed data. We validate our proposed scheme on image and text datasets. For example, our method outperforms random sampling methods by 1.24% on CIFAR-100 using Wide-ResNet-28-10. Furthermore, we achieve comparable accuracy to the SoTA Adversarial AutoAugment on CIFAR-10, CIFAR-100, SVHN, and ImageNet datasets.