| Log Message: |
MAHOUT-479: Implemented ClusterClassifier, ClusterPolicy(s) and ClusterIterator which can duplicate k-means and Dirichlet clustering in sequential execution only. Added unit tests and switched DisplayKMeans and DisplayDirichlet to use the ClusterIterator. Gives a pretty good sanity check of the clustering.
Changed pdf() implementation of GaussianCluster to use the product of the component pdfs vs the average. Seems to work a lot better.
Deprecated a number of old Dirichlet models and the experimental VectorModelClassifier.
Changed the type of AbstractCluster numPoints to long from int
All unit tests run
|
mahout/trunk/core/src/main/java/org/apache/mahout/clustering/AbstractCluster.java