The Complete SPRLIB & ANNLIB

parzen_dset_best_s

- compute the optimal smoothing parameter for a class

SYNOPSIS

double parzen_dset_best_s (dset, NCls, class, init_s)

ARGUMENTS

DATASET *dset The dataset to be used.
int NCls The number of classes.
int class The class number of which the smoothing parameter has to be estimated.
double init_s An initial smoothing parameter s.

RETURNS

The function returns the optimal smoothing value if within MAXITER (standard set to 100) iterations a better value is found. It returns init_s if no convergence was possible.

FUNCTION

The function determines the optimal smoothing paramter s for a parzen kernel for a particular class class given the dataset dset. It takes as input the class for which the kernelsize has to be estimated and an initial smoothing parameter init_s. The function iterates towards an optimal s, if an optimal s is possible.

SEE ALSO

find_points(1)

This document was generated using api2html on Thu Mar 5 09:00:00 MET DST 1998