depddp

density-enhanced Principal Direction Divisive Partitioning algorithm
[IDX,T] = DEPDDP(X, VARARGIN)

 [IDX, T] = DEPDDP(X, K) produces a divisive hierarchical clustering of the
 N-by-D data matrix (X) into (K) clusters. This algorithm uses a hierarchy of
 binary partitions each splitting the observations by first projecting onto
 the first principal component and then identifying the lowest local minimum
 of the 1D KDE constructed from the projected data.

  [IDX,T] = dePDDP(X,K) returns the cluster assignment, (IDX), and the  binary 
  tree (T) containing the cluster hierarchy. If K==[] the number of clusters is estimated

  [IDX, T] = dePDDP(X, K, 'PARAM1',val1, 'PARAM2',val2, ...) specifies optional parameters
  in the form of Name,Value pairs. 

  'bandwidth' - Bandwidth parameter
    Function Handle: bandwidth(X,pars) returns bandwidth (positive scalar)
    (default: bandwidth =0.9* std(projections) * N^(-0.2))

  'split_index' - Criterion determining which cluster to split    next (only relevant if K is specified)
    Function Handle: index = split_index(v, X, pars)
            (v: projection vector, X:data matrix, pars: parameters structure)
    Cluster with MAXIMUM INDEX is split at each step of the algorithm
    Two standard choices of split index can be enabled by setting 'split_index' to 
    one of the strings below:
        + 'size':    Split largest cluster
        + 'fval':    Split cluster whose local minimum on the 1D KDE is the lowest
    (default: split_index = 'fval')

  'minsize' - Minimum cluster size (integer)
    (default minsize = 1)

  'labels' - true cluster labels. Only used for performance assessment.

Reference:
S.K. Tasoulis, D.K. Tasoulis and V.P. Plagianakos. Enhancing principal direction divisive clustering.
Pattern Recognition, 43(10):3391-3411, 2010.