How to Do Multi-way Classification with Two-Way Classifiers

  • Florin Cutzu
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 2714)


A new principle for performing polychotomous classification with pairwise classifiers is introduced: if pairwise classifier 375-01, trained to discriminate between classes i and j, responds “i” for an input x from an unknown class (not necessarily i or j), one can at best conclude that x ∉. Thus, the output of pairwise classifier 375-02 can be interpreted as a vote against the losing class j, and not, as existing methods propose, as a vote for the winning class i. Both a discrete and a continuous classification model derived from this principle are introduced.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Erin L. Allwein, Robert E. Schapire, and Yoram Singer. Reducing multiclass to binary: A unifying approach for margin classifiers. In Proc. 17th International Conf. on Machine Learning, pages 9–16. Morgan Kaufmann, San Francisco, CA, 2000.Google Scholar
  2. 2.
    Jerome H. Friedman. Another approach to polychotomous classification. Technical report, Stanford University, 1996.Google Scholar
  3. 3.
    Trevor Hastie and Robert Tibshirani. Classification by pairwise coupling. In Michael I. Jordan, Michael J. Kearns, and Sara A. Solla, editors, Advances in Neural Information Processing Systems, volume 10. The MIT Press, 1998.Google Scholar
  4. 4.
    Eddy Mayoraz and Ethem Alpaydin. Support vector machines for multi-class classification. In IWANN (2), pages 833–842, 1999.Google Scholar
  5. 5.
    Volker Roth. Probabilistic discriminative kernel classifiers for multi-class problems. Lecture Notes in Computer Science, 2191:246–266, 2001.CrossRefGoogle Scholar
  6. 6.
    Jürgen Schürmann. Patern Classification. A Unified View of Statistical and Neural Principles. John Wiley & Sons, Inc, New York, NA, 1996.Google Scholar
  7. 7.
    B. Zadrozny. Reducing multiclass to binary by coupling probability estimates, 2001.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2003

Authors and Affiliations

  • Florin Cutzu
    • 1
  1. 1.Dept. of Computer ScienceIndiana UniversityBloomingtonUSA

Personalised recommendations