A Method for Identifying Differences between Two Classes.

Abstract

A non-parametric measure of the difference between sample distributions of a random variable for two classes of data is presented. The method involves counting the number of class reversals among the ordered set of two class data and provides a normalized measure of class intermingling. Applications of the method to the discrimination-feature selection problem are described. (Author)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 29, 1978
Accession Number
ADA064724

Entities

People

  • Nathan E. Lindgren

Organizations

  • Massachusetts Institute of Technology

Tags

DTIC Thesaurus Topics

  • Calibration
  • Cell Count
  • Computer Programming
  • Databases
  • Dimensionality Reduction
  • Feature Extraction
  • Feature Selection
  • Gaussian Distributions
  • Heart Rate
  • Information Science
  • Machine Learning
  • Mathematical Analysis
  • Mathematics
  • Measurement
  • Pattern Recognition
  • Probability
  • Random Variables

Readers

  • Computer Vision.
  • Graph Algorithms and Convex Optimization.
  • Mathematical Modeling and Probability Theory.