Thank you, Jake! I appreciate the effort. This is going to be one of those classic "30 minutes from commit to deployment" cases :) I'll be waiting for the new code, and will let you know how it works right after. Best, Danny
Hi danny, The hadoopified version of the lanczos impl currently in mahout is really easy, but not ported in from decomposer yet. There were some other things I wanted to get ready first, but users priorities trump long term plans! I'll get that over this week. I will also be adding stochastic decomposition soon after. Regarding the scaling of what is in there now, however, I should say that the stream-oriented GHA svd impl, it isn't hadoop, but it scales to 10's of millions of rows on one box, but does require a nice stream oriented matrix impl (also coming this week). -jake -- View this message in context: http://old.nabble.com/PCA-with-Mahout-tp27423079p27435128.html Sent from the Mahout User List mailing list archive at Nabble.com.
