Dataset Augmentation in Feature Space

By Terrance Devries, Graham W. Taylor. 2017

460 citations recorded by the source corpus.

Dataset augmentation, the practice of applying a wide array of domain-specific transformations to synthetically expand a training set, is a standard tool in supervised learning. While effective in tasks such as visual recognition, the set of transformations must be carefully designed, implemented, and tested for every new domain, limiting its re-use and generality. In this paper, we adopt a simpler, domain-agnostic approach to dataset augmentation. We start with existing data points and apply simple transformations such as adding noise, interpolating, or extrapolating between them. Our main insight is to perform the transformation not in input space, but in a learned feature space. A re-kindling of interest in unsupervised representation learning makes this technique timely and more effective. It is a simple proposal, but to-date one that has not been tested empirically. Working in the space of context vectors generated by sequence-to-sequence models, we demonstrate a technique that is effective for both static and sequential data.

Subfields: Artificial Intelligence, Cancer Research, 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 Cancer Research

In Representation & Self-Supervised Learning

Citations in the corpus

Randaugment: Practical automated data augmentation with a reduced search space (2019, methodology): “Data augmentation is a necessary method for achieving state-of-the-art performance [36, 17, 6, 48, 11, 30].”
mixup: Beyond Empirical Risk Minimization (2017, methodology): “More related to mixup , DeVries & Taylor (2017) show that interpolation and extrapolation in feature space can improve generalization.”
AutoAugment: Learning Augmentation Policies from Data (2018, methodology): “DeVries and Taylor used simple transformations in the learned feature space to augment data [43].”
Automatically Learning Data Augmentation Policies for Dialogue Tasks (2019, background): “There has been extensive work that employs data augmentation in both computer vision (Simard et al., 2003; Krizhevsky et al., 2012; Cireşan et al., 2012; Wan et al., 2013; Sato et al., 2015; DeVries and Taylor, 2017; Tran et al., 2017; Lemley et al., 2017) and NLP (Fürstenau and Lapata, 2009; Sennri”
Greedy AutoAugment (2019, background): “Another example is the usage of data augmentation in unsupervised learning [6], [7], [8], [9].”
Population Based Augmentation: Efficient Learning of Augmentation Policy Schedules (2019, methodology): “These include ShearX/Y, TranslateX/Y, Rotate, AutoContrast, Invert, Equalize, Solarize, Posterize, Contrast, Color, Brightness, Sharpness, Cutout (Devries & Taylor, 2017), and Sample-Pairing (Inoue, 2018).”
Learning data augmentation policies using augmented random search (2018, background): “In [12], the author proposed to deploy the augmentation in feature space.”
Conditional variance penalties and domain shift robustness (2017, background): “2017; Devries and Taylor 2017).”
Online Continual Learning Via Candidates Voting (2021, methodology): “Therefore, we adopt random perturbation for feature augmentation [10].”
Learning to Compose Domain-Specific Transformations for Data Augmentation (2017, methodology): “For example, the well-known SMOTE algorithm applies this basic technique for oversampling in class-imbalanced settings [2], and recent work explores using a similar interpolation approach in a learned feature space [5].”
Shakedrop Regularization for Deep Residual Learning (2018, background): “1), in the forward pass, Shake-Shake interpolates the outputs of two residual branches (i.e., F1(x) and F2(x)) with random weight α. DeVries and Taylor [2] demonstrated that the interpolation of two data in the feature space can synthesize reasonable augmented data; hence the interpolation in the fo”
Learning Data Augmentation Strategies for Object Detection (2019, methodology): “Recent work has shown that instead of manually designing data augmentation strategies, learning an optimal policy from data can lead to significant improvements in generalization performance of image classification models [22, 45, 8, 33, 31, 54, 2, 43, 37, 5].”
Data Augmentation with Manifold Exploring Geometric Transformations for Increased Performance and Robustness (2019, methodology): “Approaches range from elastic transformations [2], noise generation in a learned features space [3], to repeat, rotate and inﬁll approaches whereby a known sample is scaled and rotated in a grid pattern, and background consistency is ensured [4].”
Evaluating Gender Bias in Natural Language Inference (2021, background): “Data augmentation have been shown to be remarkably effective in achieving robust generalization performance in computer vision (DeVries & Taylor, 2017) as well as NLP (Andreas, 2020b).”
Image Data Augmentation for Deep Learning: A Survey (2022, background): “In [DeVries and Taylor, 2017a], authors claimed that when traversing along the manifold it is more likely to encounter realistic samples in feature space than compared to input space.”
Guarding Against Adversarial Domain Shifts with Counterfactual Regularization (2017, background): “m data and the focus often lies on geometric transformations of the input data or explicitly creating augmented observations (Sohn and Lee, 2012; Khasanova and Frossard, 2017; Hashimoto et al., 2017; Devries and Taylor, 2017). The main dierence between this line of work and CoRe is that we do not re”
Cross-Domain Few-Shot Classification via Learned Feature-Wise Transformation (2020, methodology): “Unlike the hand-crafted approaches such as horizontal ﬂip-ping and random cropping, several recent approaches have been proposed to learn the data augmentation (Cubuk et al., 2019; DeVries & Taylor, 2017a; Lemley et al., 2017; Perez & Wang, 2017; Sixt et al., 2018; Tran et al., 2017).”