LINEAR DISCRIMINANT-ANALYSIS FOR 2 CLASSES VIA REMOVAL OF CLASSIFICATION STRUCTURE

A new method for two-class linear discriminant analysis, called ''remo val of classification structure,'' is proposed. Its novelty lies in th e transformation of the data along an identified discriminant directio n into data without discriminant information and iteration to obtain t he next discriminant direction. It is free to search for discriminant directions oblique to each other and ensures that the informative dire ctions already found will not be chosen again at a later stage. The ef ficacy of the method is examined for two discriminant criteria. Studie s with a wide spectrum of synthetic data sets and a real data set indi cate that the discrimination quality of these criteria can be improved by the proposed method.