Major Description
Sometimes, the number of points passed to the reducer from the mappers in the StreamingKMeansDriver job is too large to fit into memory.
In that case, applying another StreamingKMeans pass can collapse the mapper intermediate clusters to a more manageable size to be clustered.