Abstract:
We show a method for approximating a discriminant function with a cylindrical surface in classification and image recognition problems by sequential approximations to the zero of the discriminant function (cylinder's directrix) with weighted least squares. This approach improves the efficiency of decision rules. We show characteristic examples.
Keywords:classification, image recognition, approximation of discriminant functions, weighted least squares.
Presented by the member of Editorial Board:A. I. Matasov
Fulltext:
PDF file (4583 kB)
