Semi-Supervised Classification Based on Classification from Positive and Unlabeled Data
Tomoya Sakai, Marthinus Christoffel Plessis, Gang Niu, Masashi Sugiyama
Abstract
Most of the semi-supervised classification methods developed so far use unlabeled data for regularization purposes under particular distributional assumptions such as the cluster assumption. In contrast, recently developed methods of classification from positive and unlabeled data (PU classification) use unlabeled data for risk evaluation, i.e., label information is directly extracted from unlabeled data. In this paper, we extend PU classification to also incorporate negative data and propose a novel semi-supervised learning approach. We establish generalization error bounds for our novel methods and show that the bounds decrease with respect to the number of unlabeled data without the distributional assumptions that are required in existing semi-supervised learning methods. Through experiments, we demonstrate the usefulness of the proposed methods.
BibTeX
@InProceedings{pmlr-v70-sakai17a,
title = {Semi-Supervised Classification Based on Classification from Positive and Unlabeled Data},
author = {Tomoya Sakai and Marthinus Christoffel du Plessis and Gang Niu and Masashi Sugiyama},
booktitle = {Proceedings of the 34th International Conference on Machine Learning},
pages = {2998--3006},
year = {2017},
editor = {Precup, Doina and Teh, Yee Whye},
volume = {70},
series = {Proceedings of Machine Learning Research},
month = {06--11 Aug},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v70/sakai17a/sakai17a.pdf},
url = {https://proceedings.mlr.press/v70/sakai17a.html},
abstract = {Most of the semi-supervised classification methods developed so far use unlabeled data for regularization purposes under particular distributional assumptions such as the cluster assumption. In contrast, recently developed methods of classification from positive and unlabeled data (PU classification) use unlabeled data for risk evaluation, i.e., label information is directly extracted from unlabeled data. In this paper, we extend PU classification to also incorporate negative data and propose a novel semi-supervised learning approach. We establish generalization error bounds for our novel methods and show that the bounds decrease with respect to the number of unlabeled data without the distributional assumptions that are required in existing semi-supervised learning methods. Through experiments, we demonstrate the usefulness of the proposed methods.}
}