Next: Discussion:
Up: Clustering
Previous: Algorithms:
The ``K-center'' -algorithm focuses on a radically different approach.
Instead of dividing the data into similarity groups it tries to separate a group
of K data-points from the rest of the data (2 clusters). These K records are chosen as the
most representative records within the data, which means that the
distance from any record to its closest ``representative'' is minimized.
The resulting K centers then can be used for showing the diversity of the data.
Also representative neurons for Radial Basis Functions can be created using
this algorithm.
Thomas Prang
1998-06-07