Header logo is ei

Inference with the Universum

2006

Conference Paper

ei


WIn this paper we study a new framework introduced by Vapnik (1998) and Vapnik (2006) that is an alternative capacity concept to the large margin approach. In the particular case of binary classification, we are given a set of labeled examples, and a collection of "non-examples" that do not belong to either class of interest. This collection, called the Universum, allows one to encode prior knowledge by representing meaningful concepts in the same domain as the problem at hand. We describe an algorithm to leverage the Universum by maximizing the number of observed contradictions, and show experimentally that this approach delivers accuracy improvements over using labeled data alone.

Author(s): Weston, J. and Collobert, R. and Sinz, F. and Bottou, L. and Vapnik, V.
Book Title: ICML 2006
Journal: Proceedings of the 23rd International Conference on Machine Learning (ICML 2006)
Pages: 1009-1016
Year: 2006
Month: June
Day: 0
Editors: Cohen, W. W., A. Moore
Publisher: ACM Press

Department(s): Empirical Inference
Bibtex Type: Conference Paper (inproceedings)

DOI: 10.1145/1143844.1143971
Event Name: 23rd International Conference on Machine Learning
Event Place: Pittsburgh, PA, USA

Address: New York, NY, USA
Digital: 0
Language: en
Organization: Max-Planck-Gesellschaft
School: Biologische Kybernetik

Links: PDF
Web

BibTex

@inproceedings{3916,
  title = {Inference with the Universum},
  author = {Weston, J. and Collobert, R. and Sinz, F. and Bottou, L. and Vapnik, V.},
  journal = {Proceedings of the 23rd International Conference on Machine Learning (ICML 2006)},
  booktitle = {ICML 2006},
  pages = {1009-1016},
  editors = {Cohen, W. W., A. Moore},
  publisher = {ACM Press},
  organization = {Max-Planck-Gesellschaft},
  school = {Biologische Kybernetik},
  address = {New York, NY, USA},
  month = jun,
  year = {2006},
  month_numeric = {6}
}