ON SELECTING VARIABLES FOR PATTERN CLASSIFICATION.
Abstract
The purpose of this report is two-fold. When the separating boundary between two pattern classes is a hyperplane, the effects of positive and negative correlations (between the variables used to construct the hyperplane) on the probability of misclassification are investigated. Secondly, a method for including or rejecting the variables in the hyperplane for a desired proportional gain at every stage is presented. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 19, 1967
- Accession Number
- AD0654149
Entities
People
- Poduri S. R. S. Rao