Takeru Miyato

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

2018 · IEEE Transactions on Pattern Analysis and Machine Intelligence

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 …

Robustness to Adversarial Perturbations in Learning from Incomplete Data

2019 · Neural Information Processing Systems

What is the role of unlabeled data in an inference problem, when the presumed underlying distribution is adversarially perturbed? To provide a concrete answer to this question, this paper unifies two major learning frameworks: Semi-Supervised …