You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mahout.apache.org by Stefano Bellasio <st...@gmail.com> on 2011/01/17 20:28:35 UTC
Steps to build a recommendation system with Mahout and SVD
Hi guys, i will try to have some informations always on topic SVD and recommendations, here my goals:
1) using Mahout to compute a data set as Movielens with SVD to produce recommendations
2) Using that in a cluster with hadoop
My question is, can you describe or list the steps to build up something? Do i need to start from distributedLanczosSolver or anything else?
Thank you
Re: Steps to build a recommendation system with Mahout and SVD
Posted by Sean Owen <sr...@gmail.com>.
There is no distributed SVD recommender per se.
There is a distributed SVD -- dig around the DistributedLanczosSolver code
and such. (I'm not the expert there.)
There is a basic non-distributed recommender based on the SVD in
SVDRecommender though.
On Mon, Jan 17, 2011 at 7:28 PM, Stefano Bellasio <stefanobellasio@gmail.com
> wrote:
> Hi guys, i will try to have some informations always on topic SVD and
> recommendations, here my goals:
>
> 1) using Mahout to compute a data set as Movielens with SVD to produce
> recommendations
> 2) Using that in a cluster with hadoop
>
> My question is, can you describe or list the steps to build up something?
> Do i need to start from distributedLanczosSolver or anything else?
>
> Thank you
Re: Steps to build a recommendation system with Mahout and SVD
Posted by Sebastian Schelter <ss...@apache.org>.
I started doing some work on a distributed recommender that uses matrix
factorization in https://issues.apache.org/jira/browse/MAHOUT-542
It's far away from being usable in production systems but it might be
interesting for you if you're looking at it for research purposes.
--sebastian
On 17.01.2011 20:28, Stefano Bellasio wrote:
> Hi guys, i will try to have some informations always on topic SVD and recommendations, here my goals:
>
> 1) using Mahout to compute a data set as Movielens with SVD to produce recommendations
> 2) Using that in a cluster with hadoop
>
> My question is, can you describe or list the steps to build up something? Do i need to start from distributedLanczosSolver or anything else?
>
> Thank you