Share Email Print
cover

Proceedings Paper

Small-sample error estimation: mythology versus mathematics
Author(s): Ulisses Braga-Neto
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

Error estimation is a key aspect of statistical pattern recognition. The true classification error rate is usually unavailable since it depends on the unknown feature-label distribution. Hence, one needs to estimate the error rate from the available sample data. This paper presents a concise, mathematically rigorous review of the subject of error estimation in statistical pattern recognition, pointing to the pitfalls that arise in small-sample settings due to the use of "rules of thumb" and a neglect for proper mathematical understanding of the problem.

Paper Details

Date Published: 30 August 2005
PDF: 11 pages
Proc. SPIE 5916, Mathematical Methods in Pattern and Image Analysis, 59160V (30 August 2005); doi: 10.1117/12.619331
Show Author Affiliations
Ulisses Braga-Neto, Aggeu Magalhaes Research Ctr. (Brazil)


Published in SPIE Proceedings Vol. 5916:
Mathematical Methods in Pattern and Image Analysis
Jaakko T. Astola; Ioan Tabus; Junior Barrera, Editor(s)

© SPIE. Terms of Use
Back to Top