You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mahout.apache.org by pranay venkata <sv...@gmail.com> on 2010/06/10 22:53:47 UTC
SVD algorithm
Hi,
Can SVDRecommender be used on binary preference ratings for producing
recommendations ?
--
Thanks in advance,
Pranay.
Re: SVD algorithm
Posted by André Panisson <pa...@di.unito.it>.
Hi, just to give my 2 cents about SVD Recommender. It is a model-based
recommender, and requires training before calculating recommendations,
but is very fast when recommending, differently from item/user-based
recommenders, that require no training but require more resources when
recommending. The training step computational complexity depends on the
number of user feedbacks: the current implementation is good for very
sparse user/item matrices. Important to say here that, in this case, a
matrix is considered sparse not if it has high number of zeros, but it
should have high number of unknown values. So, if your binary preference
ratings matrix has a rating 0/1 for every pair user/value, you should
consider it not as a binary rating, but as an implicit feedback. For a
good discussion about implicit feedbacks I recommend reading Koren's
paper "Collaborative Filtering for Implicit Feedback Datasets" available
at http://research.yahoo.com/pub/2433.
Regards,
André Panisson
On 06/11/2010 04:37 PM, Sean Owen wrote:
> Depending on your data, I'd imagine simple item-based recommenders
> will be faster, that's all. The SVD takes some time to compute.
>
> On Fri, Jun 11, 2010 at 3:28 PM, Nishant Chandra
> <ni...@gmail.com> wrote:
>
>> AFAIK, SVD is to overcome sparsity in the user - item matrix. How is it
>> connected to efficiency here? Do you foresee any problem?
>>
>>
>
--
André Panisson, Ph.D. Student
panisson@di.unito.it
Dipartimento di Informatica - Universita' di Torino
Corso Svizzera, 185
10149, Torino, Italy
tel. (++39.)011.6706833
fax. (++39.)011.751603
http://www.di.unito.it/~panisson
Re: SVD algorithm
Posted by Sean Owen <sr...@gmail.com>.
Depending on your data, I'd imagine simple item-based recommenders
will be faster, that's all. The SVD takes some time to compute.
On Fri, Jun 11, 2010 at 3:28 PM, Nishant Chandra
<ni...@gmail.com> wrote:
> AFAIK, SVD is to overcome sparsity in the user - item matrix. How is it
> connected to efficiency here? Do you foresee any problem?
>
Re: SVD algorithm
Posted by Nishant Chandra <ni...@gmail.com>.
AFAIK, SVD is to overcome sparsity in the user - item matrix. How is it
connected to efficiency here? Do you foresee any problem?
Thanks,
Nishant
evaluate the result -- would be interesting to hear your results.
However I'll also say I don't imagine this is the most efficient
recommender for this kind of data.
On Thu, Jun 10, 2010 at 9:53 PM, pranay venkata <sv...@gmail.com> wrote:
> Hi,
>
> Can SVDRecommender be used on binary preference ratings for producing
> recommendations ?
>
> --
> Thanks in advance,
> Pranay.
>
Re: SVD algorithm
Posted by Sean Owen <sr...@gmail.com>.
I have not tried it myself. In principle I do not see a reason you
couldn't send in vectors with 0 or 1 values only. You would have to
evaluate the result -- would be interesting to hear your results.
However I'll also say I don't imagine this is the most efficient
recommender for this kind of data.
On Thu, Jun 10, 2010 at 9:53 PM, pranay venkata <sv...@gmail.com> wrote:
> Hi,
>
> Can SVDRecommender be used on binary preference ratings for producing
> recommendations ?
>
> --
> Thanks in advance,
> Pranay.
>