[jira] Commented: (MAHOUT-493) Explicit filter for user/item pairs in o.a.m.cf.taste.hadoop.item.RecommenderJob

["Sebastian Schelter (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MAHOUT-493?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12906817#action_12906817 ] 

Sebastian Schelter commented on MAHOUT-493:
-------------------------------------------

I did not like the idea to load a file containing all user/item pairs to filter into memory in the final Reducer which was proposed on the mailing list, as this might introduce scaling issues.

I think I found a better way to include the filtering logic in the distributed recommendation process. For a user that should not be recommended an item we join an artificial NaN summand into the estimation of his/her preference towards that item so that this item will automatically be excluded.

Patch is coming soon.

> Explicit filter for user/item pairs in o.a.m.cf.taste.hadoop.item.RecommenderJob
> --------------------------------------------------------------------------------
>
>                 Key: MAHOUT-493
>                 URL: https://issues.apache.org/jira/browse/MAHOUT-493
>             Project: Mahout
>          Issue Type: New Feature
>          Components: Collaborative Filtering
>    Affects Versions: 0.5
>            Reporter: Sebastian Schelter
>
> It should be possible to explicitly prevent having a given item being recommended to a given user.
> This filter is necessary because that user might have stated to never be shown a recommended item again.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.