Some dual problems in pattern recognition

摘要

Duality concepts in pattern recognition research are described. Measurements are considered to form the coordinate axes of the primal space, and sample patterns are considered to provide the coordinate axes of the dual space. Two examples are given of algorithms for the specification of linear discriminants, and the application of duality concepts to these algorithms is described. A brief description of the analogy between cluster analysis and factor analysis is also given in pattern recognition terminology.