Class-Specific Feature Sets in Classification

Abstract

The classical Bayesian approach to classification requires knowledge of the probability density function (PDF) of the data or sufficient statistic under all class hypotheses. Because it is difficult or impossible to obtain a single low-dimensional sufficient statistic, it is often necessary to utilize a sub-optimal yet still relatively high-dimensional feature set. The performance of such an approach is severely limited by the ability to estimate the PDF on a high-dimensional space training data. A new theorem shows that by introducing a special 'noise-only' signal class (HO), it is possible to re-formulate the classical approach based upon M sufficient statistics, one corresponding to each signal class. Furthermore, the optimal classifier requires knowledge of only the PDF's of the sufficient statistics under the corresponding signal class and under noise-only condition. We present simulation results of a 9-class synthetic problem showing dramatic improvements over the traditional high-dimensional approach.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Sep 01, 1998
Accession Number
ADA477363

Entities

People

  • Paul Baggenstoss

Organizations

  • Naval Undersea Warfare Center

Tags

DTIC Thesaurus Topics

  • Applied Computer Science
  • Artificial Intelligence
  • Artificial Intelligence Computing
  • Artificial Intelligence Software
  • Bayesian Networks
  • Computational Processes
  • Computational Science
  • Computer Science
  • Computing-Related Activities
  • Data Science
  • Information Operations
  • Information Science
  • Intelligent Systems
  • Machine Learning
  • Probability
  • Probability Density Functions
  • Statistics

Readers

  • Neural Network Machine Learning.
  • Statistical inference.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Machine Learning Algorithms
  • Space
  • Space - Space Objects