Regularization with Latent Space Virtual Adversarial Training
This work offers an incremental improvement in regularization techniques for image classification, primarily benefiting researchers and practitioners working with semi-supervised learning and adversarial training.
The paper proposes Latent Space Virtual Adversarial Training (LVAT) to improve the effectiveness of Virtual Adversarial Training (VAT) by injecting perturbations in the latent space rather than the input space. This approach allows for more flexible generation of adversarial samples, leading to more effective regularization and outperforming VAT and other state-of-the-art methods on SVHN and CIFAR-10 datasets in image classification tasks.
Virtual Adversarial Training (VAT) has shown impressive results among recently developed regularization methods called consistency regularization. VAT utilizes adversarial samples, generated by injecting perturbation in the input space, for training and thereby enhances the generalization ability of a classifier. However, such adversarial samples can be generated only within a very small area around the input data point, which limits the adversarial effectiveness of such samples. To address this problem we propose LVAT (Latent space VAT), which injects perturbation in the latent space instead of the input space. LVAT can generate adversarial samples flexibly, resulting in more adverse effects and thus more effective regularization. The latent space is built by a generative model, and in this paper, we examine two different type of models: variational auto-encoder and normalizing flow, specifically Glow. We evaluated the performance of our method in both supervised and semi-supervised learning scenarios for an image classification task using SVHN and CIFAR-10 datasets. In our evaluation, we found that our method outperforms VAT and other state-of-the-art methods.