Let a set of objects exist each of which is described by N features X1? ..., XN, where each feature X} is a real number. So each object is set by N-dimensional vector (Xl5 ..., XN) and represents a point in the space of object descriptions, RN.There are also set objects for which degrees of membership in either class are unknown. A decision rule should be determined that could enable estimation of the membership of either object with unknown degrees of membership in the given classes (Ozols and Borisov, 1996). To determine the decision rule, such features should be found which give a possibility to distinguish objects belonging to different classes, i.e. features that are specific for each class. That is why a subtask of estimation of the efficiency of features should be solved. A function 5 should be determined which could enable estimation of the efficiency of both separate features and of features groups.Thus, the task is reduced to the determination of a number of features from set N that will best describe groups of objects and will enable possibly correct recognition of the object's membership in a class.
A note on sequential decision approach to pattern recognition and machine learning
