Nahh... the names are the key. On Tue, May 24, 2011 at 2:49 PM, Jeff Eastman <jeast...@narus.com> wrote:
> If the D vectors are NamedVectors, with their index as the name, then this > will flow through to the clustered points at the output. The order of those > points won't bear much relation to the order of the input, but the names > will be preserved. KMeans does not mess with the order of the elements > within each D vector. I don't know if this is sufficient or if Lanczos does > anything similar.