Adversarial AutoAugment

By Qiang Wang, Jian Zhang, Zhaobai Zhong. 2019

215 citations recorded by the source corpus.

Data augmentation (DA) has been widely utilized to improve generalization in training deep neural networks. Recently, human-designed data augmentation has been gradually replaced by automatically learned augmentation policy. Through finding the best policy in well-designed search space of data augmentation, AutoAugment can significantly improve validation accuracy on image classification tasks. However, this approach is not computationally practical for large-scale problems. In this paper, we develop an adversarial method to arrive at a computationally-affordable solution called Adversarial AutoAugment, which can simultaneously optimize target related object and augmentation policy search loss. The augmentation policy network attempts to increase the training loss of a target network through generating adversarial augmentation policies, while the target network can learn more robust features from harder examples to improve the generalization. In contrast to prior work, we reuse the computation in target network training for policy evaluation, and dispense with the retraining of the target network. Compared to AutoAugment, this leads to about 12x reduction in computing cost and 11x shortening in time overhead on ImageNet. We show experimental results of our approach on CIFAR-10/CIFAR-100, ImageNet, and demonstrate significant performance improvements over state-of-the-art. On CIFAR-10, we achieve a top-1 test error of 1.36%, which is the currently best performing single model. On ImageNet, we achieve a leading performance of top-1 accuracy 79.40% on ResNet-50 and 80.00% on ResNet-50-D without extra data.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Model Architectures, Safety, Robustness & Interpretability

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 Computer Vision and Pattern Recognition

In Model Architectures

In Safety, Robustness & Interpretability

Citations in the corpus

A Comprehensive Survey of Neural Architecture Search (2020, background): “…also quickly applied in the fields of object detection [65, 75, 111, 118], semantic segmentation [63, 64, 120], adversarial learning [53], architectural scaling [114, 122, 124], multi-objective optimization [39, 115, 125], platform-aware [28, 34, 103, 117], data augmentation [121, 123] and so on.”
UniformAugment: A Search-free Probabilistic Data Augmentation Approach (2020, background): “Another interesting direction for further work is to compare UA’s eﬀective-ness against approaches using adversarial methods for policy generation [19].”
Image Augmentations for GAN Training (2020, background): “The augmentation strategies also range from the basic operations like random crop and horizontal ﬂip to more sophisticated handcrafted operations [10, 39, 43, 16], or even the strategies directly learned by the neural network [8, 44].”
MaxUp: A Simple Way to Improve Generalization of Neural Network Training (2020, unknown): “More recently, more advanced methods have been developed to ﬁnd the optimal data augmentation policies using reinforcement learning or adversarial generative network (e.g. Cubuk et al., 2019a;b; Zhang et al., 2020).”
On the Generalization Effects of Linear Transformations in Data Augmentation (2020, methodology): “For image classiﬁcation tasks, we compare Algorithm 1 with AutoAugment (AA) (Cubuk et al., 2018), Fast AutoAugment (Fast AA) (Lim et al., 2019), Population Based Augmentation (PBA) (Ho et al., 2019), RandAugment (RA) (Cubuk et al., 2019), and Adversarial AutoAugment (Adv.”
Self-supervised Learning of Adversarial Example: Towards Good Generalizations for Deepfake Detection (2022, methodology): “Ours w/ ran ops); the second uses a same synthesizer network to select different operations and magnitudes for different samples, which is similar to the setting in [55] (i.”
A Comprehensive Survey of Image Augmentation Techniques for Deep Learning (2022, background): “Adversarial learning-based ADA [69], CDST-DA [70], Ada-Transform [71], Adversarial AA [72], IF-DA [73], SPA [74].”
Circumventing Outliers of AutoAugment with Knowledge Distillation (2020, background): “Researchers provided comments from two aspects: some of them argued that the transformation strategies may have been overly aggressive and need to be controlled [17], while some others advocated for the benefit of exploring aggressive transformations so that richer information is integrated into the”
Time Series Data Augmentation for Deep Learning: A Survey (2020, methodology): “The idea of automated data augmentation is to automatically search for optimal data augmentation policies through re-inforcement learning, meta learning, or evolutionary search [Ratner et al. , 2017; Cubuk et al. , 2019; Zhang et al. , 2020; Cheung and Yeung, 2021 ] .”
Semi-Supervised and Unsupervised Deep Visual Learning: A Survey (2022, background): “to manually design domain-specific data augmentation [76], [77], [83], [84], [85].”
TrivialAugment: Tuning-free Yet State-of-the-Art Data Augmentation (2021, unknown): “AA) [22] is another slightly cheaper method that uses multiple workers and learns the augmentation policy online.”
SmoothLLM: Defending Large Language Models Against Jailbreaking Attacks (2023, background): “38 [81] Xinyu Zhang, Qiang Wang, Jian Zhang, and Zhao Zhong.”