DATA REDUCTION USING INFORMATION THEORETIC TECHNIQUES.
Abstract
Final results are presented from a mathematical study to improve the overall performance of pattern recognition devices, such as for reconnaissance photograph interpretation. The specific results of this study are: (1) the design of practical pattern recognition systems requires mathematical procedures to account for statistical interdependencies and redundancy. (2) Inadequate experimental data sets are a common situation in practice and quantitative error formulas from sampling theory should be employed to assess and monitor the inadequacy. (3) A data reduction and combining technique was developed and tested, which accounts for the redundancy and data inadequacy. (4) A method of evaluating competing recognizer designs was developed from statistical sampling theory.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1967
- Accession Number
- AD0813395
Entities
People
- G. F. Hughes
- J. A. Lebo