Share Email Print

Proceedings Paper

Classifier performance model for the many-class case
Author(s): William J. Schonlau
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

A model is presented for predicting classification performance for systems having a large population of classes. The cases of large and small training set size for each class are treated separately. A method is proposed for measuring classification performance as the mean ranking statistic (rho) E which is derived from the average information content hE of the system feature vector, which is in turn derived from the system covariance matrices ((Sigma) W, (Sigma) B). This method for predicting (rho) E is applied to the large training set case (case 1), explaining why performance is not compromised but improved by adding noisy features. The method is extended for predicting performance in the more difficult small training set case (case 2), explaining why performance may be compromised by the addition of noisy features in that situation.

Paper Details

Date Published: 7 August 2002
PDF: 9 pages
Proc. SPIE 4728, Signal and Data Processing of Small Targets 2002, (7 August 2002);
Show Author Affiliations
William J. Schonlau, Raytheon Electronic Systems (United States)

Published in SPIE Proceedings Vol. 4728:
Signal and Data Processing of Small Targets 2002
Oliver E. Drummond, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?