Printable Version

Back to Previous Page

Iterative Least Squares Linear Decision Surface

This command develops a linear discriminant using an iterative least squares algorithm. The weight vector will oscillate for as long as it is allowed, either by the maximum number of iterations, or until linear separation is achieved. The method makes use of the hyperbolic tangent function as an approximation to the binary response of a two-class problem. This allows the use of derivative calculations to speed the convergence of the separation. Also, the use of an internal table of function values allows more rapid calculations.