SDKNN k-nearest neighbor classifier

    P=SDKNN(DATA,options)

 INPUT
   DATA      training dataset

 OPTIONS
   k         number of neighbors (def: 1)
   proto     number of prototypes to select per class (def: [] = use all samples)
   protosel  prototype selection method (def: 'random')
              'kcentres','random'
   method    method to compute k-NN with k>1 (def: 'kappa')
              'kappa' - outputs per-class distance to k-th neighbor (one/multi class)
              'classfrac' - outputs class fraction between k neighbors (only multi class)
 OUTPUT
   P         pipeline object

 DESCRIPTION
 SDKNN trains a k-NN classifier (by default k=1). One- or multi-class k-NN
 are supported. By default all provided examples are used as
 prototypes. Prototype selection may be performed by setting number of
 desired prototypes using the 'proto' option.

sdknn is referenced in examples:

• kb18: How to protect a trained discriminant against outliers?
• kb16: Visualize the effect of a change of parameters in a trained classifier