Different Topic: There's a talk this afternoon by Cloudera's Data Scientist at MlConf NYC about Mahout's LanczosSolver, SSVD and MlLib SSVD.
See http://mlconf.com/mlconf-2015-nyc/ Here we r talking about purging Mahout's LanczosSolver for 2 years now. Seems like the talk will be about the old MapReduce based SSVD and LanczosSolver while we have the new non-MR distributed SSVD stuff. I hope I am wrong here but will see. On Fri, Mar 27, 2015 at 1:02 PM, Shannon Quinn <[email protected]> wrote: > Honestly not sure, as I haven't had a chance to play around with the scala > dsl much yet. Suneel suggested we save that for 0.10.1. > > > On 3/27/15 12:00 PM, Dmitriy Lyubimov wrote: > >> Shannon, >> >> How difficult would it be to port spectral clustering to our scala alg and >> math? We have ssvd there as well. >> On Mar 27, 2015 7:26 AM, "Shannon Quinn (JIRA)" <[email protected]> wrote: >> >> Shannon Quinn created MAHOUT-1659: >>> ------------------------------------- >>> >>> Summary: Remove deprecated Lanczos solver from spectral >>> clustering in mr-legacy >>> Key: MAHOUT-1659 >>> URL: https://issues.apache.org/jira/browse/MAHOUT-1659 >>> Project: Mahout >>> Issue Type: Task >>> Components: Clustering, mrlegacy >>> Affects Versions: 0.9 >>> Reporter: Shannon Quinn >>> Assignee: Shannon Quinn >>> Priority: Minor >>> Fix For: 0.10.0 >>> >>> >>> Spectral clustering still has the option of using either SSVD or the >>> Lanczos solver for dimensionality reduction. Remove the latter entirely. >>> >>> >>> >>> -- >>> This message was sent by Atlassian JIRA >>> (v6.3.4#6332) >>> >>> >
