Accession Number:

AD0742283

Title:

A Multivariate Data Display and Analysis Procedure

Descriptive Note:

Technical rept.

Corporate Author:

OHIO STATE UNIV COLUMBUS ELECTROSCIENCE LAB

Personal Author(s):

• Breeding, Kenneth J.

Report Date:

1972-04-01

Pagination or Media Count:

31.0

Abstract:

The paper seeks to extend the conventional linear, and non-linear, feature space mappings used in pattern recognition. Of particular interest when data are highly interdependent, the minimal spanning trees of Zahn are shown to contain a great deal of information about the distribution of this type of data. Further, trees are shown to be indicative of the quantization level necessary for application of nearest neighbor recognition schemes. A linear projection is developed to display the data and confirm distributions applicable to the minimal spanning trees. Author

Descriptors:

• *PATTERN RECOGNITION
• VECTOR ANALYSIS
• INFORMATION THEORY
• MAPPING(TRANSFORMATIONS)

Subject Categories:

• Cybernetics

Distribution Statement:

APPROVED FOR PUBLIC RELEASE