THE PERCEPTRON CORRECTION PROCEDURE IN NON-SEPARABLE SITUATIONS,

Abstract

The behavior of the standard Perceptron correction procedure when the underlying patterns are non-separable is discussed. It is shown that in this case the magnitude of the succeeding weight vectors remains bounded. Moreover, if the magnitudes approach a limit this limit is actually attained, and the ensuing behavior is particularly simple. A converse to the usual Perceptron convergence theorem is presented. (Author)

Document Details

Document Type
Technical Report
Publication Date
Feb 01, 1964
Accession Number
AD0433144

Entities

People

  • Bradley Efron

Organizations

  • SRI International

Tags

DTIC Thesaurus Topics

  • Convergence

Readers

  • Calculus or Mathematical Analysis
  • Mathematical Modeling and Probability Theory.
  • Neural Network Machine Learning.