This is the transpose of how things are done. The distributed row matrix has wide patches of a matrix that contain all of a few rows (hence the name). If you need columns, you have to do the algebra to figure out how to operate on the transpose of your matrix so that you can store it they way that Mahout does.
On Wed, Sep 28, 2011 at 10:07 AM, Marc Sturlese <[email protected]>wrote: > I've thought maybe it possible to write a seqfile with rows from 1 to 10 > and > columns from 1 to 10. Then another seqfile with rows from 1 to 10 (again) > but now columns are from 11 to 20. Is this format understood by a > DistributedRowMatrix if I pass to it as path the folder where all this > seqfiles are contained? > Thanks in advance > > -- > View this message in context: > http://lucene.472066.n3.nabble.com/about-DistributedRowMatrix-implementation-tp3375372p3375596.html > Sent from the Mahout User List mailing list archive at Nabble.com. >
