Virtual Adversarial Training

Making Attention Mechanisms More Robust and Interpretable with Virtual Adversarial Training for Semi-Supervised Text Classification

We propose a new general training technique for attention mechanisms based on virtual adversarial training (VAT). VAT can compute adversarial perturbations from unlabeled data in a semi-supervised setting for the attention mechanisms that have been …

半教師あり文書分類のための仮想敵対的学習による注意機構の頑健性および解釈性の向上

言語処理学会第 27 回年次大会,2021.