[Spark ML] existence of Matrix Factorization ALS algorithm's log version

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

[Spark ML] existence of Matrix Factorization ALS algorithm's log version

James Yuan
Hi Team,

I'm looking for information regarding MF_ALS algorithm's log version if
implemented. In original Hu et al.'s paper "Collaborative filtering for
implicit feedback datasets", there're two different ways to construct
confidence level c(ui) from raw info r(ui): the vanilla version is
c(ui)=1+alpha*r(ui), the log version is c(ui)=1+alpha*log(1+rui/epsilon).
I'm guessing the default version in pyspark.ml.recommedation.ALS is vanilla
version. Not sure if we have log version available or not.

Any help is appreciated :)

James



--
Sent from: http://apache-spark-user-list.1001560.n3.nabble.com/

---------------------------------------------------------------------
To unsubscribe e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: [Spark ML] existence of Matrix Factorization ALS algorithm's log version

srowen
No there isn't a log version. You could probably copy and hack the
implementation easily if necessary.

On Wed, Jul 29, 2020 at 11:05 AM jyuan1986 <[hidden email]> wrote:

>
> Hi Team,
>
> I'm looking for information regarding MF_ALS algorithm's log version if
> implemented. In original Hu et al.'s paper "Collaborative filtering for
> implicit feedback datasets", there're two different ways to construct
> confidence level c(ui) from raw info r(ui): the vanilla version is
> c(ui)=1+alpha*r(ui), the log version is c(ui)=1+alpha*log(1+rui/epsilon).
> I'm guessing the default version in pyspark.ml.recommedation.ALS is vanilla
> version. Not sure if we have log version available or not.
>
> Any help is appreciated :)
>
> James
>
>
>
> --
> Sent from: http://apache-spark-user-list.1001560.n3.nabble.com/
>
> ---------------------------------------------------------------------
> To unsubscribe e-mail: [hidden email]
>

---------------------------------------------------------------------
To unsubscribe e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: [Spark ML] existence of Matrix Factorization ALS algorithm's log version

James Yuan
Thanks for your quick reply. I'll hack it if needed :)

James



--
Sent from: http://apache-spark-user-list.1001560.n3.nabble.com/

---------------------------------------------------------------------
To unsubscribe e-mail: [hidden email]