PARZENML

Optimum smoothing parameter in Parzen density estimation.

    H = PARZENML(A)

Input
A	Input dataset

Output
H	Scalar smoothing parameter (in case of crisp labels)  Vector with smoothing parameters (in case of soft labels)

Description

Maximum likelihood estimation for the smoothing parameter H in the  Parzen denstity estimation of the data in A. A leave-one out  maximum likelihood estimation is used.

The optimisation of H may be stopped prematurely by PRTIME.

The dataset A can either be crisp or soft labeled. In case of crisp  labeling the class information is not used and a single smoothing  parameter is estimated. In case of soft labels a smoothing parameter  for every class is estimated and objects are weighted in relation to  their class weigthts (soft label value).  It may be profitable to scale the data before calling it. eg.  WS = SCALEM(A,'variance'); A = A*WS.

See also

datasets, mappings, scalem, seldat, parzenm, parzendc, prprogress,