Spectral Clustering

Assemble unknown data into cluster

spcl(data, nbclusters, varargin) is a spectral clustering function to assemble random unknown data into clusters. after specifying the data and the number of clusters, next parameters can vary as wanted. This function will construct the fully connected similarity graph of the data. The first parameter of varargin is the name of the function to use, the second is the parameter to pass to the function.
Third parameter is the type of the Laplacian matrix:
'unormalized' - unnormalized laplacian matrix
'sym' - normalized symmetric laplacian matrix
'rw' - normalized asymmetric laplacian matrix
(if omitted the default will be 'unnormalized')
then the algorithm used for organizing eigenvectors:
'np' - generally used for 2 clusters, one eigenvector must be used, if will put positive values in class 1 and negative values in class 2
'kmean' - a k-mean algorithm will be used to cluster the given eigenvectors

finally an eigenvector choice can be added, it can be a vector [vmin vmax] or a matrix defining several intervals. if not found the default will be [2 2]