[
https://issues.apache.org/jira/browse/MAHOUT-586?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12982944#action_12982944
]
Lance Norskog edited comment on MAHOUT-586 at 1/17/11 8:18 PM:
---------------------------------------------------------------
Sure, this is not intended as a committable patch. It's more a statement of
philosophy- and yeah, the first version of RecommenderEvaluator should stay.
The problem being solved is that I'm messing around with Recommenders and
derived DataModels and wanted to measure them against the existing ones, and a
modular design helped.
bq. More importantly this is also removing all of the logic to hold out test
data.
Holding out test data is done with the subsampling data model.
bq. GroupLensRecommenderEvaluatorRunner does nothing but invoke an evalution;
it's like 5 lines of code. So I'm not sure I understand the comment.
The patch has a new GLRER at the end that uses the modular code to mimic the
old class.
bq. there is no root-mean-square evaluation anymore?
RMS can be done with another of the RunningAverage classes, right? Thank you
for pointing this out.
was (Author: lancenorskog):
Sure, this is not intended as a committable patch. It's more a statement of
philosophy- and yeah, the first version of RecommenderEvaluator should stay.
The problem being solved is that I'm messing around with Recommenders and
derived DataModels and wanted to measure them against the existing ones, and a
modular design helped.
bq. More importantly this is also removing all of the logic to hold out test
data.
Holding out test data is done with the subsampling data model.
bq. GroupLensRecommenderEvaluatorRunner does nothing but invoke an evalution;
it's like 5 lines of code. So I'm not sure I understand the comment.
The patch has a new GLRER that uses the modular code to effect the old class.
bq. there is no root-mean-square evaluation anymore?
RMS can be done with another of the RunningAverage classes, right? Thank you
for pointing this out.
> 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
> Attachments: MAHOUT-586.patch
>
>
> The RecommenderEvaluator implementation is hard-coded around one algorithm.
> This is a more flexible, modular rewrite.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
