×

Self-organizing neural network for pattern classification

  • US 5,682,503 A
  • Filed: 10/11/1994
  • Issued: 10/28/1997
  • Est. Priority Date: 02/12/1991
  • Status: Expired due to Fees
First Claim
Patent Images

1. A neural network system for classifying an input pattern, comprising:

  • a) means for comparing the input pattern to templates of already learned patterns, each template having a class associated with said template;

    b) means for selecting a class for the input pattern based on results of the comparison by the means for comparing;

    c) means for determining whether a correct class was selected by the means for selecting for the input pattern;

    d) means, operative in response to a determination that a correct class was selected for the input pattern, for determining whether the correct class has been stably classified by comparing a template of the selected class which is most similar to the input pattern to a template of another class which is most similar to the input pattern among templates of said another class to determine whether the compared templates are similar to a degree that classification may not be accurate; and

    e) means, operative in response to a determination that the correct class has not been stably classified, for adjusting the compared templates to ensure stable classification.

View all claims
  • 0 Assignments
Timeline View
Assignment View
    ×
    ×