[
https://issues.apache.org/jira/browse/MAHOUT-586?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13010566#comment-13010566
]
Lance Norskog edited comment on MAHOUT-586 at 3/24/11 5:54 AM:
---------------------------------------------------------------
(The MAHOUT-586 of March 15 has the SamplingDataModel classes)
I was working with a datamodel that did its own recommendations.
The original preference-based api (AbstractDifference)RecommenderEvaluator api
does not allow for comparing a data model with a recommender or another data
model. AbstractDifference... with a little refactoring does this using its
existing sampling code. I've sent my version of this to a couple of people on
the list, because they wanted to evaluate a datamodel against something. If you
adopt this refactoring, the api becomes much more usable. Also much faster than
my rearrangement :)
Second, I was working with a datamodel that did not give numbers at all.
Order-based seemed the right way, but far too many algorithms presented
themselves.
The problem with this patch is that I have not posted these recommenders, so it
all seems pointless. But believe me you, there's a reason for this
pointlessness :)
was (Author: lancenorskog):
(The MAHOUT-586 of March 15 has the SamplingDataModel classes)
I was working with a datamodel that did its own recommendations.
The original preference-based api (AbstractDifference)RecommenderEvaluator api
does not allow for comparing a data model with a recommender or another data
model. AbstractDifference... with a little refactoring does this using its
existing sampling code. I've sent my version of this to a couple of people on
the list, because they wanted to evaluate a datamodel against something. If you
adopt this refactoring, the api becomes much more usable. Also much faster than
my rearrangement :)
Second, I was working with a datamodel that did not give numbers at all.
Order-based seemed the right way, but far too many algorithms presented
themselves.
The problem with this patch is that I have not posted these recommenders, so it
all seems pointless. But believe me you, there's a reason for my pointlessness
:)
> 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: 0.5
>
> Attachments: MAHOUT-559.patch, MAHOUT-586.patch, MAHOUT-586.patch,
> 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.
For more information on JIRA, see: http://www.atlassian.com/software/jira
