[ 
https://issues.apache.org/jira/browse/MAHOUT-586?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Lance Norskog updated MAHOUT-586:
---------------------------------

    Fix Version/s:     (was: 0.5)
                   0.6
      Description: 
The RecommenderEvaluator implementation is hard-coded around the standard 
algorithm for comparing a recommender's performance on training v.s. test 
datasets.

This is a rewrite supplying other algorithms, and the ability to directly 
compare DataModels. It was born out of a desire to check that a new recommender 
made sense compared to existing recommenders, and to directly work with 
DataModels that encompass the entire recommendation space.


  was:
The RecommenderEvaluator implementation is hard-coded around one algorithm.
This is a more flexible, modular rewrite.


I'm happy with this iteration. It is trimmed down to three implementations that 
make sense:
* Precision (directly matching the research paper meaning)
* Mean Reciprocal Rank (creatively interpreted)
* Bubble sort (should be Ulam or Levenshtein distance)

And includes the ability to track the average or RMS of these comparisons.

Please reopen and commit this.

> Redo RecommenderEvaluator for modularity
> ----------------------------------------
>
>                 Key: MAHOUT-586
>                 URL: https://issues.apache.org/jira/browse/MAHOUT-586
>             Project: Mahout
>          Issue Type: Improvement
>          Components: Collaborative Filtering
>            Reporter: Lance Norskog
>            Assignee: Sean Owen
>             Fix For: 0.6
>
>         Attachments: MAHOUT-559.patch, MAHOUT-586.patch, MAHOUT-586.patch, 
> MAHOUT-586.patch, MAHOUT-586.patch, MAHOUT-586.patch
>
>
> The RecommenderEvaluator implementation is hard-coded around the standard 
> algorithm for comparing a recommender's performance on training v.s. test 
> datasets.
> This is a rewrite supplying other algorithms, and the ability to directly 
> compare DataModels. It was born out of a desire to check that a new 
> recommender made sense compared to existing recommenders, and to directly 
> work with DataModels that encompass the entire recommendation space.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to