Active GPU capabilities in regression algorithms (not only dl4j)

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

Active GPU capabilities in regression algorithms (not only dl4j)

forky
Is it possible to activate the use of the GPU using common ML techniques on
Weka? I've seen this option only for dl4j



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: Active GPU capabilities in regression algorithms (not only dl4j)

Eibe Frank
I have used the GPU-based implementation of XGBoost for CUDA (made at Waikato!) through the RPlugin (or the wekaPython wrapper, not entirely sure?).

The MTJ matrix algebra package used by some WEKA algorithms is based on netlib-java, which, in theory, should make it possible to use cuBLAS or a similar library as a backend, see


I would love to hear if you can make this work.

Cheers,
Eibe

Some relevant info from an older post:

WEKA implementations of algorithms that are based on standard linear algebra generally apply the MTJ library, which, optionally, can use a much faster native backend. Install the appropriate netlibNative* package for your platform using the WEKA package manager to get this acceleration. To get further speed-ups, compile OpenBLAS or similar for your particular computer and link it with MTJ/NetlibJava.To see really big speed improvements for WEKA schemes that use matrix-based computations, you need to use such a system-optimized native library. There is good news for Mac OS X users: OS X comes with a system-optimized library (vecLib) and all you need to do is install the WEKA package netlibNativeOSX to get a very nice speed boost.

Here is a list of schemes (not sure whether it’s still complete) that can benefit from this (but only on sufficiently large data):

GaussianProcesses
PrincipalComponents
LinearRegression
M5P
M5Rules
MultivariateGaussianEstimator

There are also some schemes in various packages:

LDA
QDA
FLDA
LatentSemanticAnalysis
Nystroem
RotationForest
LeastMedSq
RBFNetwork

If you want to explore other options, you can also use the RPlugin or the wekaPython packages (and maybe even wekaPyScript) to run algorithms available via R or Python in WEKA.


On Thu, Mar 4, 2021 at 10:37 PM forky <[hidden email]> wrote:
Is it possible to activate the use of the GPU using common ML techniques on
Weka? I've seen this option only for dl4j



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html