Capacity control in linear classifiers for pattern recognition

I. Guyon, V. Vapnik, B. Boser, L. Bottou, Sara A Solla

Research output: Chapter in Book/Report/Conference proceedingConference contribution

10 Scopus citations

Abstract

Achieving good performance in statistical pattern recognition requires matching the capacity of the classifier to the amount of training data. If the classifier has too many adjustable parameters (large capacity), it is likely to learn the training data without difficulty, but will probably not generalize properly to patterns that do not belong to the training set. Conversely, if the capacity of the classifier is not large enough, it might not be able to learn the task at all. In between, there is an optimal classifier capacity which ensures the best expected generalization for a given amount of training data. The method of Structural Risk Minimization (SRM) refers to tuning the capacity of the classifier to the available amount of training data. In this paper, we illustrate the method of SRM with several examples of algorithms. We present experiments which confirm theoretical predictions of performance improvement in application to handwritten digit recognition.

Original languageEnglish (US)
Title of host publicationConference B
Subtitle of host publicationPattern Recognition Methodology and Systems
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages385-388
Number of pages4
ISBN (Print)0818629150
DOIs
StatePublished - Jan 1 1992
Event11th IAPR International Conference on Pattern Recognition, IAPR 1992 - The Hague, Netherlands
Duration: Aug 30 1992Sep 3 1992

Publication series

NameProceedings - International Conference on Pattern Recognition
Volume2
ISSN (Print)1051-4651

Other

Other11th IAPR International Conference on Pattern Recognition, IAPR 1992
CountryNetherlands
CityThe Hague
Period8/30/929/3/92

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition

Fingerprint Dive into the research topics of 'Capacity control in linear classifiers for pattern recognition'. Together they form a unique fingerprint.

  • Cite this

    Guyon, I., Vapnik, V., Boser, B., Bottou, L., & Solla, S. A. (1992). Capacity control in linear classifiers for pattern recognition. In Conference B: Pattern Recognition Methodology and Systems (pp. 385-388). [201798] (Proceedings - International Conference on Pattern Recognition; Vol. 2). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICPR.1992.201798