Cheat-sheetSDKMEANS k-means classifier or clustering P=SDKMEANS(DATA,options) P=SDKMEANS(DATA,K,options) INPUT DATA training dataset K number of clusters per class OPTIONS 'k' number of clusters per class, may be vector (required) 'all' execute k-means on entire data set (not per class) 'iter' max.number of iterations (opt, def:200) 'cluster' return one output per cluster (default: return one output per class=classifier) 'nodisplay' do not show any output 'no pruning' do not prune the k-means classifier (returns all trained prototypes but yields lower performance) 'prefix' Custom cluster name prefix (default: 'C') 'k final' k in the final k-NN classifier (default: 1) OUTPUT P pipeline object DESCRIPTION SDKMEANS describes data by means of k clusters. The clusters are defined by an iterative algorithm started from randomly selected samples. Iterations stop when no samples change their cluster membership or at max.number of iterations defined by 'iter' option. By default SDKMEANS trains a 1-NN classifier which handles each class in DATA separately and returns one output per class (square Euclidean distance to the closest cluster). By default, per-class prototypes are pruned removing the ones causing errors on the training set. Use 'no pruning' option to return all per-class prototypes. The k in the final k-NN classifier may be changed with 'k final' option. Data clustering may be performed using 'cluster' option. SDKMEANS then returns one output per cluster. The number of centers may be specified using 'k' parameter (vector of 'k', one per class is supported') READ MORE http://perclass.com/doc/guide/clustering.html#sdkmeans SEE ALSO SDKNN, SDKCENTRES, SDCLUSTER, SDMIXTURE
sdkmeans is referenced in examples: