perClass Documentation
version 5.1 (31-May-2017)
 SDPARZEN Parzen density

     P=SDPARZEN(DATA,options)

 INPUT
   DATA     SDDATA set or data matrix

 OUTPUT
   P        Parzen pipeline

 OPTIONS
   vector         use vector smoothing parameter
   h              smoothing parameter ('scalar' or 'vector')
   iter           number of iterations (def: [] = use maximum smooting
                  difference delta to stop)
   delta          maximum smoothing difference (def: 1e-6)
   maxsamples     limit max number of samples used (default: use all)
   prior          class priors (default: use priors from the training set)
   gpu            run training on GPU (requires parallel comp.toolbox)

 DESCRIPTION
 SDPARZEN implements non-parametric Parzen classifier. In training, it
 estimates smoothing parameter using EM algorithm. By default, Laplace
 kernel and scalar smoothing parameter is used. Scalar and vector
 smoothing parameters are supported.  Estimation is stoped when delta
 difference on likelihood is reached. Alternatively, user may specify
 fixed number of iterations.

 READ MORE
http://perclass.com/doc/guide/classifiers.html#sdparzen

sdparzen is referenced in examples: