That information is discarded in order to make the streaming job possible. To recreate this information, you would need to do another pass over the data in order to assign records to the closest centroid. There may already be such a job, but it is not likely to be compatible with the streaming k-means.
Making these pieces fit (or writing them) would be an excellent project and I am sure that the community would help you in doing so. On Fri, Jul 5, 2013 at 12:42 PM, sirvan paraste <[email protected]>wrote: > Hi, > I asked this question last week, > > In generated output for streamingkmenas , there is now information of > elements of each clusters, > I think it should be implemented, please guide me, what I must do? I am new > to mahout > > thanks, >
