You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mahout.apache.org by Grant Ingersoll <gs...@apache.org> on 2009/03/18 18:30:34 UTC

Netflix Demo for Taste

Hi,

I see there is a Netflix recommender for Taste checked in.  Is there  
directions some where for running it?  Something akin to the Group  
Lens instructions would be pretty cool.

Seems like we could package up a JAR just like group lens, but I'm not  
sure about where to put the data, etc.

-Grant

Re: Netflix Demo for Taste

Posted by Grant Ingersoll <gs...@apache.org>.
On Mar 18, 2009, at 4:23 PM, Sean Owen wrote:

> Would be easy to set up: needs a Recommender with a no-arg  
> constructor that
> contains all the setup logic. Then I think the target that builds
> 'grouplens.jar' should then build all the examples. Then the user  
> could just
> write in a different Recommender.

Yeah, I was thinking we could do this w/ several of the Taste  
examples, and just have the JAR be taste-examples.jar

>
>
> The Netflix data set is much much larger: the code has to choose to  
> use only
> about 1% of the data or else the memory requirement passes a  
> gigabyte, which
> is probably too much for a casual demo. It isn't really a  
> recommender that
> uses the whole data set, so maybe that is why I didn't set it up as  
> a proper
> demo.

OK

>
>
> There is more code in there which was bits of a more sophisticated  
> effort
> which is suitable for running in batch on a big machine - not demo  
> friendly.

Good to know

>
>
>
> Sean
>
> On Mar 18, 2009 5:31 PM, "Grant Ingersoll" <gs...@apache.org>  
> wrote:
>
> Hi,
>
> I see there is a Netflix recommender for Taste checked in.  Is there
> directions some where for running it?  Something akin to the Group  
> Lens
> instructions would be pretty cool.
>
> Seems like we could package up a JAR just like group lens, but I'm  
> not sure
> about where to put the data, etc.
>
> -Grant


Re: Netflix Demo for Taste

Posted by Sean Owen <sr...@gmail.com>.
Would be easy to set up: needs a Recommender with a no-arg constructor that
contains all the setup logic. Then I think the target that builds
'grouplens.jar' should then build all the examples. Then the user could just
write in a different Recommender.

The Netflix data set is much much larger: the code has to choose to use only
about 1% of the data or else the memory requirement passes a gigabyte, which
is probably too much for a casual demo. It isn't really a recommender that
uses the whole data set, so maybe that is why I didn't set it up as a proper
demo.

There is more code in there which was bits of a more sophisticated effort
which is suitable for running in batch on a big machine - not demo friendly.


Sean

On Mar 18, 2009 5:31 PM, "Grant Ingersoll" <gs...@apache.org> wrote:

Hi,

I see there is a Netflix recommender for Taste checked in.  Is there
directions some where for running it?  Something akin to the Group Lens
instructions would be pretty cool.

Seems like we could package up a JAR just like group lens, but I'm not sure
about where to put the data, etc.

-Grant