Open AccessPosted Content
Learning to Compose Domain-Specific Transformations for Data Augmentation
TL;DR: In this article, a generative adversarial approach is proposed to learn a sequence model over user-specified transformation functions using GANs, which can make use of arbitrary, non-deterministic transformation functions, and is robust to misspecified user input.
read more
Abstract: Data augmentation is a ubiquitous technique for increasing the size of labeled training sets by leveraging task-specific data transformations that preserve class labels. While it is often easy for domain experts to specify individual transformations, constructing and tuning the more sophisticated compositions typically needed to achieve state-of-the-art results is a time-consuming manual task in practice. We propose a method for automating this process by learning a generative sequence model over user-specified transformation functions using a generative adversarial approach. Our method can make use of arbitrary, non-deterministic transformation functions, is robust to misspecified user input, and is trained on unlabeled data. The learned transformation model can then be used to perform data augmentation for any end discriminative model. In our experiments, we show the efficacy of our approach on both image and text datasets, achieving improvements of 4.0 accuracy points on CIFAR-10, 1.4 F1 points on the ACE relation extraction task, and 3.4 accuracy points when using domain-specific transformation operations on a medical imaging dataset as compared to standard heuristic augmentation approaches.
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
Figures
Figure 1: Three examples of transformation functions (TFs) in different domains: Two example sequences of incremental image TFs applied to CIFAR-10 images (left); a conditional word-swap TF using an externally trained language model and specifically targeting nouns (NN) between entity mentions (E1,E2) for a relation extraction task (middle); and an unsupervised segementation-based translation TF applied to mass-containing mammography images (right). 
Figure 7: Accuracy scores on random 10% subsamples of test data (dotted lines) and on versions augmented with a single transformation (vertical bars) with parameters drawn uniformly at random. 
Figure 2: A high-level diagram of our method. Users input a set of transformation functions h1, ..., hK and unlabeled data. A generative adversarial approach is then used to train a null class discriminator, D∅, and a generator, G, which produces TF sequences hτ1 , ..., hτL . Finally, the trained generator is used to perform data augmentation for an end discriminative model Df . 
Table 4: A simple study of the effect of adding a transformation regularization (TR) term to the objective function, evaluated on a labeled validation set. We see that adding the term improves performance for both heuristic (random) TF sequences and for TF sequences generated by the trained LSTM model, and that there is a larger positive effect for the latter. 
Table 3: Current state-of-the-art image classification models as ranked by reported performance on the CIFAR-10 and CIFAR-100 tasks, and their error with (Err. w/ DA) and without (Err. w/o DA) data augmentation. We include both scores and particular data augmentation techniques when reported, although the latter is rarely reported with great precision. 13 
Table 1: Test set performance of end models trained on subsamples of the labeled training data (% ), not including validation splits, using various data augmentation approaches. None indicates performance with no augmentation. All tasks are measured in accuracy, except ACE which is measured by F1 score.
Citations
AutoAugment: Learning Augmentation Strategies From Data
Ekin D. Cubuk,Barret Zoph,Dandelion Mane,Vijay K. Vasudevan,Quoc V. Le +4 more
- 15 Jun 2019
TL;DR: This paper describes a simple procedure called AutoAugment to automatically search for improved data augmentation policies, which achieves state-of-the-art accuracy on CIFAR-10, CIFar-100, SVHN, and ImageNet (without additional data).
•Proceedings Article
RandAugment: Practical Automated Data Augmentation with a Reduced Search Space
Ekin D. Cubuk,Barret Zoph,Jonathon Shlens,Quoc V. Le +3 more
- 01 Jan 2020
TL;DR: This work proposes a simplified search space that vastly reduces the computational expense of automated augmentation, and permits the removal of a separate proxy task.
Randaugment: Practical automated data augmentation with a reduced search space
Ekin D. Cubuk,Barret Zoph,Jonathon Shlens,Quoc V. Le +3 more
- 14 Jun 2020
TL;DR: EfficientNet-B7 as mentioned in this paper proposes a simplified search space that greatly reduces the computational expense of automated augmentation, and permits the removal of a separate proxy task, and achieves state-of-the-art results in image classification and object detection.
Albumentations: fast and flexible image augmentations
Alexander Buslaev,Vladimir Iglovikov,Eugene Khvedchenya,Alex Parinov,Mikhail Druzhinin,Alexandr A. Kalinin +5 more
TL;DR: Albumentations as mentioned in this paper is a fast and flexible open source library for image augmentation with many various image transform operations available that is also an easy-to-use wrapper around other augmentation libraries.
1.8K
•Posted Content
AutoAugment: Learning Augmentation Policies from Data
TL;DR: This paper describes a simple procedure called AutoAugment to automatically search for improved data augmentation policies, which achieves state-of-the-art accuracy on CIFAR-10, CIFar-100, SVHN, and ImageNet (without additional data).
1.6K
References
•Posted Content
Deep Residual Learning for Image Recognition
TL;DR: This work presents a residual learning framework to ease the training of networks that are substantially deeper than those used previously, and provides comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth.
117.9K
Gradient-based learning applied to document recognition
Yann LeCun,Léon Bottou,Léon Bottou,Yoshua Bengio,Yoshua Bengio,Yoshua Bengio,Patrick Haffner,Patrick Haffner +7 more
- 01 Jan 2001
TL;DR: This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task, and Convolutional neural networks are shown to outperform all other techniques.
32.7K
•Posted Content
Explaining and Harnessing Adversarial Examples
TL;DR: The authors argue that the primary cause of neural networks' vulnerability to adversarial perturbation is their linear nature, which is supported by new quantitative results while giving the first explanation of the most intriguing fact about adversarial examples: their generalization across architectures and training sets.
15.9K
•Posted Content
Improved Techniques for Training GANs
TL;DR: In this article, the authors present a variety of new architectural features and training procedures that apply to the generative adversarial networks (GANs) framework and achieve state-of-the-art results in semi-supervised classification on MNIST, CIFAR-10 and SVHN.
7.4K
•Posted Content
Enriching Word Vectors with Subword Information
TL;DR: A new approach based on the skipgram model, where each word is represented as a bag of character n-grams, with words being represented as the sum of these representations, which achieves state-of-the-art performance on word similarity and analogy tasks.