Application of Recognition Theory to Missile Identification and Decoy Discrimination. Volume 3. Optimum Decision Boundaries.
Abstract
Statistical classification in n-dimensional space consists in partitioning the space into category regions with decisi n boundaries and assigning an unknown to the category in whose region it falls. The wide utility is demonstrated of a particular form of decisio boundary, the hyper phere, which, while especially easy to implement, is fully optimum for large classes of distributions which may arise in real problems. Of the broad spectrum of distri utions described for which the hypersphere is optimum, particular interest centers on the normal, extended-t, and m-paraboloid distributions; and methods for obtaining the boundary parameters are prescribed. Ordering of the c ordinate directions according to their relative significance in contributing to the decision is examined, therby indicating the most efficient reduction of dimensionality where this may be desired in order to allow further computational simplicity. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 12, 1962
- Accession Number
- AD0286554
Entities
People
- Paul W. Cooper
Organizations
- Melpar