[
https://issues.apache.org/jira/browse/MAHOUT-586?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13008918#comment-13008918
]
Sean Owen commented on MAHOUT-586:
----------------------------------
My main issue with this patch is that it is deprecating the current API, but
not replacing its functionality as far as I can see, when the functionality it
provides is the more basic and essential function: computing familiar metrics
like RMSE.
It also references SamplingDataModel which doesn't exist.
I have attached my version of your latest patch which polishes up style, so
would iterate from that one.
I am not sure I agree that there is a big API problem to be solved. You're
replacing subclassing with a switch statement. I personally find it more
understandable to have subclasses which each take DataModel / Recommender as
input, and compute some metric and return it. In this API, the result comes out
by an out-parameter.
My personal opinion is just my opinion, though I think if we're going to
replace a fine and functional API with another one, there's a burden of showing
the replacement is substantially better and I'm not seeing that along the lines
this is going yet.
But I think it's good to invite others to comment?
> 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
