Hi,
I try to make it clear:
I have some Training data from the past. Then I have a Matrix Q which looks like a cube.
The Matrix have values e.g. from 1 to 3.
But some of the fields of this Matrix are empty (or zero)
Then I want to make a tensor factorization to get 3 matrix (A,B,C).
So I can get the following AxBxC = Q (not exactly but nearly Q)
I'm a student and need this for a Work that I write.
Was this helpful?
Ursprüngliche Nachricht
Von: Ted Dunning [mailto:ted.dunning@gmail.com]
Gesendet: Samstag, 16. März 2013 01:56
An: user@mahout.apache.org
Betreff: Re: Tensor SVD possible?
Can you give at least a tiny amount of details here?
Why? What is the goal? What data? What source? Time varying? Real application or historical
data? Is this an academic project?
Geographically constrained?
This is like asking about databases and saying that you "want to store some data". If you
don't give more details, then you won't get any reasonable answers.
On Fri, Mar 15, 2013 at 3:15 PM, B.Maciag < bartholomaeus.maciag@tudortmund.de> wrote:
> Hi,
>
> I want to create recommendations with it.
>
> Ted Dunning <ted.dunning@gmail.com> schrieb:
>
> >No.
> >
> >But tensor SVD isn't even uniquely defined.
> >
> >You can get similar results using some tricks, but it depends on
> >special conditions in your problem.
> >
> >Can you say what you are trying to do?
> >
> >
> >On Fri, Mar 15, 2013 at 1:31 AM, Bartholomäus Maciag <
> >bartholomaeus.maciag@tudortmund.de> wrote:
> >
> >> Hi, is it possible to do a Tensor SVD with Mahout?
> >>
> >> If yes, how?
> >>
> >>
> >>
> >> Best, B.
> >>
> >>
> >>
> >>
>
