answered Feb 6 '14
Imho this is correct behavior, since the cluster()-function calls kMeans, and kMeans is initialized randomly, i.e. the output clusters will differ.
