CONTROLLING SELECTIVITY IN NONSTANDARD PATTERN-RECOGNITION ALGORITHMS

In pattern recognition it is important to be able to control the selectivity of the classification algorithms. It is well-known that the number of objects assigned to a significant class depends on the method used. In this paper a type of aggregation operators, referred to as mixed connective, is used to summarize the information about objects to be classified (supplied by the descriptors). Because mixed connectives depend on a parameter this leads to the concept of families for these operators. Therefore, given such a family, it is possible to associate different classifications with the same data set, depending on the value chosen for the parameter, and the manner in which the algorithm selectivity classifications are compared an illustrated by an example involving a quantitative data basis.