Abstract
Purchase PDF (273 K)
E-mail Article
Add to my Quick Links
Cited By in Scopus (2)
Purchase PDF (578 K)
doi:10.1016/j.dam.2004.04.003 
Copyright © 2004 Elsevier B.V. All rights reserved.
On data classification by iterative linear partitioning
Martin Anthony
Received 5 January 2003;
revised 17 April 2004;
accepted 24 April 2004.
Available online 4 August 2004.
References and further reading may be available for this article. To view references and further reading you must purchase this article.
Abstract
We analyze theoretically the generalization properties of multi-class data classification techniques that are based on iteratively partitioning the data points by hyperplanes. A special case is that in which the data points of different classes are separated by a number of parallel hyperplanes, and we investigate the algorithmics of finding a suitable partitioning in this case.
Keywords: Pattern classification; Multi-class classification; Hyperplanes; Decision lists
Article Outline
- 1. Introduction
- 2. Threshold decision lists
- 2.1. Decision lists
- 2.2. Threshold functions and threshold decision lists
- 2.3. Multithreshold functions
- 3. Generalization from random data
- 4. Proofs of the generalization error bounds
- 5. Consistent hypothesis finders for multithreshold functions
- 5.1. Finding consistent monotonic multithreshold functions
- 5.2. Using linear programming
- 5.3. An incremental procedure
- 5.4. Some open questions in the non-monotonic case
- Acknowledgements
- References
