On the Generalization Effects of Linear Transformations in Data Augmentation

Abstract

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.

arXiv comment22 pages. Appeared in ICML 2020

References in corpus (14)

• mixup: Beyond Empirical Risk Minimization
• Unsupervised Data Augmentation for Consistency Training
• AutoAugment: Learning Augmentation Policies from Data
• Benign Overfitting in Linear Regression
• Semi-Supervised Learning with Generative Adversarial Networks
• QANet: Combining Local Convolution with Global Self-Attention for Reading Comprehension
• Data Augmentation by Pairing Samples for Images Classification
• Shake-Shake regularization
• ShakeDrop Regularization for Deep Residual Learning
• A Kernel Theory of Modern Data Augmentation
• The generalization error of max-margin linear classifiers: Benign overfitting and high dimensional asymptotics in the overparametrized regime
• Learning Data Manipulation for Augmentation and Weighting
• Understanding and Mitigating the Tradeoff Between Robustness and Accuracy
• Efficient Augmentation via Data Subsampling

Cited by in corpus (3)

• Models Genesis
• Scalable Multitask Learning Using Gradient-based Estimation of Task Affinity
• Boosting Multitask Learning on Graphs through Higher-Order Task Affinities