Jump to ContentJump to Main Navigation
Semi-Supervised Learning$
Users without a subscription are not able to see the full content.

Olivier Chapelle, Bernhard Scholkopf, and Alexander Zien

Print publication date: 2006

Print ISBN-13: 9780262033589

Published to MIT Press Scholarship Online: August 2013

DOI: 10.7551/mitpress/9780262033589.001.0001

Show Summary Details
Page of

PRINTED FROM MIT PRESS SCHOLARSHIP ONLINE (www.mitpress.universitypressscholarship.com). (c) Copyright The MIT Press, 2021. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in MITSO for personal use.date: 18 June 2021

Risks of Semi-Supervised Learning: How Unlabeled Data Can Degrade Performance of Generative Classifiers

Risks of Semi-Supervised Learning: How Unlabeled Data Can Degrade Performance of Generative Classifiers

Chapter:
(p.56) (p.57) 4 Risks of Semi-Supervised Learning: How Unlabeled Data Can Degrade Performance of Generative Classifiers
Source:
Semi-Supervised Learning
Author(s):

Cozman Fabio

Cohen Ira

Publisher:
The MIT Press
DOI:10.7551/mitpress/9780262033589.003.0004

This chapter presents a number of conclusions. Firstly, labeled and unlabeled data contribute to a reduction in variance in semi-supervised learning under maximum-likelihood estimation. Secondly, when the model is “correct,” maximum-likelihood methods are asymptotically unbiased both with labeled and unlabeled data. Thirdly, when the model is “incorrect,” there may be different asymptotic biases for different values of λ. Asymptotic classification error may also vary with λ—an increase in the number of unlabeled samples may lead to a larger estimation asymptotic bias and to a larger classification error. If the performance obtained from a given set of labeled data is better than the performance with infinitely many unlabeled samples, then at some point the addition of unlabeled data must decrease performance.

Keywords:   semi-supervised learning, maximum-likelihood estimation, asymptotic classification error, estimation asymptotic bias, unlabeled data

MIT Press Scholarship Online requires a subscription or purchase to access the full text of books within the service. Public users can however freely search the site and view the abstracts and keywords for each book and chapter.

Please, subscribe or login to access full text content.

If you think you should have access to this title, please contact your librarian.

To troubleshoot, please check our FAQs, and if you can't find the answer there, please contact us.