LatentAugment: Dynamically Optimized Latent Probabilities of Data Augmentation

TL;DR

LatentAugment introduces a dynamic, latent-variable-based approach to optimize data augmentation policies for image classification, improving accuracy across multiple datasets with theoretical and empirical validation.

Contribution

It proposes a novel latent-variable model that adaptively optimizes augmentation strategies during training, unifying and extending existing methods.

Findings

Achieves higher test accuracy on CIFAR-10, CIFAR-100, SVHN, and ImageNet.

Provides a computationally efficient and theoretically grounded augmentation framework.

Includes existing augmentation methods as special cases.

Abstract

Although data augmentation is a powerful technique for improving the performance of image classification tasks, it is difficult to identify the best augmentation policy. The optimal augmentation policy, which is the latent variable, cannot be directly observed. To address this problem, this study proposes $\textit{LatentAugment}$, which estimates the latent probability of optimal augmentation. The proposed method is appealing in that it can dynamically optimize the augmentation strategies for each input and model parameter in learning iterations. Theoretical analysis shows that LatentAugment is a general model that includes other augmentation methods as special cases, and it is simple and computationally efficient in comparison with existing augmentation methods. Experimental results show that the proposed LatentAugment has higher test accuracy than previous augmentation methods on the CIFAR-10, CIFAR-100, SVHN, and ImageNet datasets.