Honestly i don't see a big deal in keeping it around. Mllib does and nobody really cared. (They have als.train() for regular als with regularization and als.trainImplicit for the implicit one). Our primary woes with too many algorithms were associated with support, but with 2 lines it is clearly not an issue any more.
Also, assuming we are buildig ml environment, i'd rather see examples of end2end solution pipeline, using it to combine data prep and algorithms, rather than a simple thing such as als. But i don't mind putting it off into examples eventually. Just let me keep it around for the time being since it has common elements i will likely reuse (such as rmse computation pipeline). -d On May 31, 2014 11:18 PM, "Sebastian Schelter (JIRA)" <[email protected]> wrote: > > [ > https://issues.apache.org/jira/browse/MAHOUT-1566?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14014918#comment-14014918 > ] > > Sebastian Schelter commented on MAHOUT-1566: > -------------------------------------------- > > If its a mere showcase, could we maybe add it as an example in an example > package, not a full fledged algorithm implementation somehow? > > > Regular ALS factorizer with convergence test. > > --------------------------------------------- > > > > Key: MAHOUT-1566 > > URL: https://issues.apache.org/jira/browse/MAHOUT-1566 > > Project: Mahout > > Issue Type: Task > > Affects Versions: 0.9 > > Reporter: Dmitriy Lyubimov > > Assignee: Dmitriy Lyubimov > > Priority: Trivial > > Fix For: 1.0 > > > > > > ALS-related: let's start with unweighed, unregularized implementation. > > > > -- > This message was sent by Atlassian JIRA > (v6.2#6252) >
