We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:


References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Semi-Supervised Classification Based on Classification from Positive and Unlabeled Data

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 classification 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 classification methods. Through experiments, we demonstrate the usefulness of the proposed methods.
Comments: Accepted to the 34th International Conference on Machine Learning (ICML 2017)
Subjects: Machine Learning (cs.LG)
Cite as: arXiv:1605.06955 [cs.LG]
  (or arXiv:1605.06955v4 [cs.LG] for this version)

Submission history

From: Tomoya Sakai [view email]
[v1] Mon, 23 May 2016 09:37:48 GMT (153kb,D)
[v2] Fri, 14 Oct 2016 14:04:24 GMT (261kb,D)
[v3] Wed, 1 Mar 2017 11:39:31 GMT (406kb,D)
[v4] Fri, 16 Jun 2017 11:14:36 GMT (1750kb,D)

Link back to: arXiv, form interface, contact.