Re: WEKA Processing Requirements

> On Jan 22, 2023, at 7:01 PM, Peter Reutemann <[email protected]> wrote:
> 
>> I just had a few questions about the ideal hardware requirements for running WEKA. We use it quite often for processing large numbers of samples, 1000+ samples with ~400 variables. We are looking into building a custom PC to run WEKA as efficiently as possible.
>> 
>> Are there any hardware requirements we should be mindful of? Such as, would WEKA benefit from a powerful GPU? Any advice would be fantastic.
> 
> 
> 
> You can monitor your algorithm's memory consumption from the GUIChooser with:
> Program -> Memory usage
> 

I’m not familiar with this. I usually use jconsole to monitor memory usage.

I found changing the vm startup options for gc policy could make a difference in the size of the data Weka could manage.

For example…

-XX:+UseG1GC -XX:MaxGCPauseMillis=50

I discovered accidentally seemed to give a good improvement. That was a while ago something now might be better.

I have an incomplete project somewhere where I was going to try and determine optimal best practices for this.

As I remember for the classifiers I looked at performance seemed to degrade fairly linearly with data size until you were completely out of memory when it got exponentially worse, fast.

_______________________________________________
Wekalist mailing list -- [email protected]
Send posts to [email protected]
To unsubscribe send an email to [email protected]
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