Provably Learning Diverse Features in Multi-View Data with Midpoint Mixup
This addresses the feature diversity issue in classification for machine learning practitioners, offering a theoretical explanation for Mixup's success, though it is incremental as it builds on existing Mixup techniques.
The paper tackles the problem of learning multiple features per class in multi-view data, showing that while empirical risk minimization often learns only one feature, a specific Mixup instantiation successfully learns both features for every class, with theoretical and empirical validation on modified image benchmarks.
Mixup is a data augmentation technique that relies on training using random convex combinations of data points and their labels. In recent years, Mixup has become a standard primitive used in the training of state-of-the-art image classification models due to its demonstrated benefits over empirical risk minimization with regards to generalization and robustness. In this work, we try to explain some of this success from a feature learning perspective. We focus our attention on classification problems in which each class may have multiple associated features (or views) that can be used to predict the class correctly. Our main theoretical results demonstrate that, for a non-trivial class of data distributions with two features per class, training a 2-layer convolutional network using empirical risk minimization can lead to learning only one feature for almost all classes while training with a specific instantiation of Mixup succeeds in learning both features for every class. We also show empirically that these theoretical insights extend to the practical settings of image benchmarks modified to have multiple features.