Learning from Labeled and Unlabeled Data: Semi-supervised Learning and Ranking

Empirical Inference

We consider the general problem of learning from labeled and unlabeled data, which is often called semi-supervised learning or transductive inference. A principled approach to semi-supervised learning is to design a classifying function which is sufficiently smooth with respect to the intrinsic structure collectively revealed by known labeled and unlabeled points. We present a simple algorithm to obtain such a smooth solution. Our method yields encouraging experimental results on a number of classification problems and demonstrates effective use of unlabeled data.

Author(s):	Zhou, D.
Links:	PDF
Year:	2004
Month:	January
Day:	0

BibTeX Type:	Talk (talk)

Digital:	0
Electronic Archiving:	grant_archive
Event Place:	The Natural Language Computing Group of Microsoft Research Asia, and the Institute of System Sciences, the Chinese Academy of Sciences, Beijing, China
Organization:	Max-Planck-Gesellschaft
School:	Biologische Kybernetik

BibTeX

@talk{2589,
  title = {Learning from Labeled and Unlabeled Data: Semi-supervised Learning and Ranking},
  abstract = {We consider the general problem of learning from labeled and unlabeled data, which is often called semi-supervised learning or transductive inference. A principled approach to semi-supervised learning is to design a classifying function which is sufficiently smooth with respect to the intrinsic structure collectively revealed by known labeled and unlabeled points. We present a simple algorithm to obtain such a smooth solution. Our method yields encouraging experimental results on a number of classification problems and demonstrates effective use of unlabeled data.},
  organization = {Max-Planck-Gesellschaft},
  school = {Biologische Kybernetik},
  month = jan,
  year = {2004},
  author = {Zhou, D.},
  month_numeric = {1}
}