Current browse context:
stat
Change to browse by:
References & Citations
Statistics > Machine Learning
Title: Adversarial Training Methods for Semi-Supervised Text Classification
(Submitted on 25 May 2016 (v1), last revised 16 Nov 2021 (this version, v4))
Abstract: Adversarial training provides a means of regularizing supervised learning algorithms while virtual adversarial training is able to extend supervised learning algorithms to the semi-supervised setting. However, both methods require making small perturbations to numerous entries of the input vector, which is inappropriate for sparse high-dimensional inputs such as one-hot word representations. We extend adversarial and virtual adversarial training to the text domain by applying perturbations to the word embeddings in a recurrent neural network rather than to the original input itself. The proposed method achieves state of the art results on multiple benchmark semi-supervised and purely supervised tasks. We provide visualizations and analysis showing that the learned word embeddings have improved in quality and that while training, the model is less prone to overfitting. Code is available at this https URL
Submission history
From: Takeru Miyato [view email][v1] Wed, 25 May 2016 04:25:45 GMT (364kb)
[v2] Mon, 7 Nov 2016 15:59:37 GMT (370kb)
[v3] Sat, 6 May 2017 20:16:03 GMT (740kb)
[v4] Tue, 16 Nov 2021 07:16:21 GMT (1197kb)
Link back to: arXiv, form interface, contact.