edu.stanford.nlp.cluster
Class PDDP
java.lang.Object
|
+--edu.stanford.nlp.cluster.AbstractClusteringMethod
|
+--edu.stanford.nlp.cluster.PDDP
- All Implemented Interfaces:
- ClusteringMethod
- public class PDDP
- extends AbstractClusteringMethod
Principal Direction Divisive Partitioning.
PDDP (Boley, 1999) uses iterative splitting. It splits each partition by the hyperplane orthogonal to the principal direction.
Field Summary |
protected cern.colt.matrix.DoubleMatrix2D |
featurematrix
|
protected cern.colt.matrix.DoubleMatrix2D |
submatrix
|
Constructor Summary |
PDDP()
|
featurematrix
protected cern.colt.matrix.DoubleMatrix2D featurematrix
submatrix
protected cern.colt.matrix.DoubleMatrix2D submatrix
PDDP
public PDDP()
initialize
public void initialize()
split
public void split(HHCluster cluster)
- Splits
cluster
along the hyperplane orthogonal to the principal direction of the cluster. Adds both children to Clusters.
chooseNext
public HHCluster chooseNext()
- Chooses the next cluster to split.
It does this by finding the cluster with the highest scatter value.
Returns most scattered cluster
cluster
public SimpleClusters cluster(DataCollection data,
int num_clusters)
- Description copied from interface:
ClusteringMethod
- Performs clustering algorithm, and populates Clusters
Stanford NLP Group