Virtual Adversarial Training: A Regularization Method for Supervised and Semi-Supervised Learning

TL;DR

This paper introduces Virtual Adversarial Training (VAT), a novel regularization technique that enhances supervised and semi-supervised learning by smoothing model predictions around data points without requiring label information.

Contribution

The paper presents VAT, a new regularization method that improves semi-supervised learning by defining adversarial directions without label dependence, achieving state-of-the-art results.

Findings

VAT achieves state-of-the-art semi-supervised performance on SVHN and CIFAR-10.

The method requires only two forward and backward passes for gradient computation.

VAT is computationally efficient and applicable to neural networks.

Abstract

We propose a new regularization method based on virtual adversarial loss: a new measure of local smoothness of the conditional label distribution given input. Virtual adversarial loss is defined as the robustness of the conditional label distribution around each input data point against local perturbation. Unlike adversarial training, our method defines the adversarial direction without label information and is hence applicable to semi-supervised learning. Because the directions in which we smooth the model are only "virtually" adversarial, we call our method virtual adversarial training (VAT). The computational cost of VAT is relatively low. For neural networks, the approximated gradient of virtual adversarial loss can be computed with no more than two pairs of forward- and back-propagations. In our experiments, we applied VAT to supervised and semi-supervised learning tasks on multiple benchmark datasets. With a simple enhancement of the algorithm based on the entropy minimization principle, our VAT achieves state-of-the-art performance for semi-supervised learning tasks on SVHN and CIFAR-10.