Improved Mixed-Example Data Augmentation

By Cecilia Summers, Michael J. Dinneen. 2018

181 citations recorded by the source corpus.

In order to reduce overfitting, neural networks are typically trained with data augmentation, the practice of artificially generating additional training data via label-preserving transformations of existing training examples. While these types of transformations make intuitive sense, recent work has demonstrated that even non-label-preserving data augmentation can be surprisingly effective, examining this type of data augmentation through linear combinations of pairs of examples. Despite their effectiveness, little is known about why such methods work. In this work, we aim to explore a new, more generalized form of this type of data augmentation in order to determine whether such linearity is necessary. By considering this broader scope of "mixed-example data augmentation", we find a much larger space of practical augmentation techniques, including methods that improve upon previous state-of-the-art. This generalization has benefits beyond the promise of improved performance, revealing a number of types of mixed-example data augmentation that are radically different from those considered in prior work, which provides evidence that current theories for the effectiveness of such methods are incomplete and suggests that any such theory must explain a much broader phenomenon.

Subfields: Artificial Intelligence, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Representation & Self-Supervised Learning

Citations in the corpus

CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features (2019, background): “ will see in the experiments, Mixup samples suffer from the fact that they are locally ambiguous and unnatural, and therefore confuses the model, especially for localization. Recently, Mixup variants [40,34,9] have been proposed; they perform feature-level interpolation and other types of transforma”
A survey on Image Data Augmentation for Deep Learning (2019, background): “The concept of mixing images in an unintuitive way was further investigated by Summers and Dinneen [66].”
Simple is Better! Lightweight Data Augmentation for Low Resource Slot Filling and Intent Classification (2020, methodology): “rmance for SF. 5 Related Work Data augmentation methods have been widely applied in computer vision, ranging from geometric transformations (Krizhevsky et al., 2012; Zhong et al., 2020), data mixing (Summers and Dinneen, 2019) to the use of generative models (Goodfellow et al., 2014) for generating ”
FMix: Enhancing Mixed Sample Data Augmentation (2020, methodology): “Following the observation that combining interpolation and masking provides the best results, and particularly the experiments in Summers and Dinneen (2019), we considered a grey-scale version of FMix.”
Enhancing Performance of Deep Learning Models with different Data Augmentation Techniques: A Survey (2020, background): “[26] by creating more generalized form of mixing images.”
SeqMix: Augmenting Active Sequence Labeling via Sequence Mixup (2020, background): “the Mixup variants (Verma et al., 2019; Summers and Dinneen, 2019; Guo et al., 2019b) turn to perform interpolation in the hidden space to capture higher-level information.”