Performance Measures for Neutral Nets Using Johnson Distributions

Abstract

A probability distribution for multilayer perceptron artificial neural net outputs is derived assuming a sigmoidal activation function. This distribution is known to be a member of the Johnson system of distributions. Using this distribution, theoretical receiver operating characteristic curves can be developed to obtain recognition differential values for corresponding values of the probability of false alarm. Application of these techniques for the detection of broadband signals is presented.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 1993
Accession Number
ADA265102

Entities

People

  • J. T. Durham
  • R. D. Trueblood
  • W. C. Torrez

Organizations

  • Naval Command, Control and Ocean Surveillance Center

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Bayesian Networks
  • Data Science
  • Detection
  • Distribution Functions
  • False Alarms
  • Frequency
  • Gaussian Distributions
  • Goodness Of Fit Tests
  • Information Processing
  • Information Science
  • Intervals
  • Neural Networks
  • Normal Distribution
  • Probability
  • Probability Distributions
  • Random Variables
  • Signal Processing

Readers

  • Neural Network Machine Learning.
  • Radar Systems Engineering.
  • Statistical inference.