Documentation>MATLAB API>KMEANS - vl_hikmeans

[TREE,ASGN] = VL_HIKMEANS(DATA,K,NLEAVES) applies integer K-menas recursively to cluster the data DATA, returing a structure TREE representing the clusters and a vector ASGN with the data to cluster assignments. The depth of the recursive partition is computed so that at least NLEAVES are generated.

VL_HIKMEANS() is built on top of VL_IKMEANS() and requires the data to be of class UINT8.

TREE is a structure representing the hierarchical clusters. Each node of the tree is also a structure with fields:

DEPTH

Depth of the tree (only at the root node)

CENTERS

K cluster centers

SUB

Array of K node structures representing subtrees (this field is missing at leaves).

ASGN is a matrix with one column per datum and height equal to the depth of the tree. Each column encodes the branch of the tree that correspond to each datum.

Example

ASGN(:,7) = [1 5 3] means that the tree as depth equal to 3 and that the datum X(:,7) corresponds to the branch ROOT->SUB(1)->SUB(5)->SUB(3).

See also: VL_HIKMEANSPUSH(), VL_HIKMEANSHIST(), VL_IKMEANS(), VL_HELP().