[ 
https://issues.apache.org/jira/browse/MAHOUT-827?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13153762#comment-13153762
 ] 

[email protected] commented on MAHOUT-827:
------------------------------------------------------


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/2887/
-----------------------------------------------------------

Review request for mahout.


Summary
-------

RecommenderJob now supports an option called "broadcast" that determines 
whether the recommendations shall be computed with a reduce-side join (the 
current approach) or a broadcast join (new and faster approach, that is 
applicable as long as the similarity matrix fits into the memory of a mapper 
instance)


This addresses bug MAHOUT-827.
    https://issues.apache.org/jira/browse/MAHOUT-827


Diffs
-----

  
trunk/core/src/main/java/org/apache/mahout/cf/taste/hadoop/TasteHadoopUtils.java
 1204135 
  
trunk/core/src/main/java/org/apache/mahout/cf/taste/hadoop/item/AggregateAndRecommendReducer.java
 1204135 
  
trunk/core/src/main/java/org/apache/mahout/cf/taste/hadoop/item/ItemIDIndexReducer.java
 1204135 
  
trunk/core/src/main/java/org/apache/mahout/cf/taste/hadoop/item/PartialMultiplyMapper.java
 1204135 
  
trunk/core/src/main/java/org/apache/mahout/cf/taste/hadoop/item/RecommenderJob.java
 1204135 
  
trunk/core/src/main/java/org/apache/mahout/cf/taste/hadoop/item/UserVectorSplitterMapper.java
 1204135 
  
trunk/core/src/main/java/org/apache/mahout/cf/taste/hadoop/item/broadcast/Estimators.java
 PRE-CREATION 
  
trunk/core/src/main/java/org/apache/mahout/cf/taste/hadoop/item/broadcast/RecommendationsPerUserMapper.java
 PRE-CREATION 
  
trunk/core/src/main/java/org/apache/mahout/cf/taste/hadoop/item/broadcast/SimilarityMatrixIterator.java
 PRE-CREATION 
  
trunk/core/src/main/java/org/apache/mahout/cf/taste/hadoop/preparation/PreparePreferenceMatrixJob.java
 1204135 
  
trunk/core/src/main/java/org/apache/mahout/cf/taste/impl/similarity/GenericItemSimilarity.java
 1204135 
  trunk/core/src/main/java/org/apache/mahout/common/AbstractJob.java 1204135 
  
trunk/core/src/main/java/org/apache/mahout/common/iterator/sequencefile/SequenceFileDirIterator.java
 1204135 
  
trunk/core/src/main/java/org/apache/mahout/math/hadoop/similarity/cooccurrence/RowSimilarityJob.java
 1204135 
  
trunk/core/src/test/java/org/apache/mahout/cf/taste/hadoop/item/RecommenderJobTest.java
 1204135 
  
trunk/core/src/test/java/org/apache/mahout/cf/taste/hadoop/item/broadcast/SimilarityMatrixIteratorTest.java
 PRE-CREATION 
  trunk/core/src/test/java/org/apache/mahout/math/hadoop/MathHelper.java 
1204135 

Diff: https://reviews.apache.org/r/2887/diff


Testing
-------


Thanks,

Sebastian


                
> Another version of RecommenderJob that broadcasts the similarity matrix
> -----------------------------------------------------------------------
>
>                 Key: MAHOUT-827
>                 URL: https://issues.apache.org/jira/browse/MAHOUT-827
>             Project: Mahout
>          Issue Type: New Feature
>          Components: Collaborative Filtering
>    Affects Versions: 0.6
>            Reporter: Sebastian Schelter
>            Assignee: Sebastian Schelter
>         Attachments: MAHOUT-827-2.patch, MAHOUT-827-3.patch, MAHOUT-827.patch
>
>
> Add another version of RecommenderJob that computes the item similarities via 
> RowSimilarityJob but assumes that the resulting similarity matrix fits into 
> the memory of the mappers in the cluster. After the item similarity 
> computation is done, the similarities are broadcasted via Hadoop's 
> distributed cache and the recommendations are computed in a map-only pass 
> over the data afterwards.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to