Re: New WEKA release: 3.8.6/3.9.6

A note for macOS and Linux users: since sending the release announcement for WEKA 3.8.6/3.9.6, we have had to replace the release files for these two platforms on SourceForge. More specifically: 

1) The macOS DMGs have been replaced with code-signed versions so that installation of WEKA on recent macOS versions became possible.

2) Just now, the ZIP files for the Linux operating system have been replaced with new ZIP files, where a permission setting for the included Java virtual machine has been fixed.

The Zulu OpenJDK JRE included in the previous ZIP files for Linux had incorrect file permissions for the jspawnhelper file, which made it impossible to spawn processes with the exec() method. This functionality is needed by the RPlugin to install basic libraries in R, and it is also needed by Auto-WEKA to spawn separate WEKA processes.

Rather than downloading the new Linux ZIP files, you can alternatively just fix the permission settings yourself (if you have permission to do so on your Linux system that is). Assuming that you are in the directory containing the weka-3-9-6 folder or the weka-3-8-6 folder obtained from the corresponding ZIP file, you can run:

chmod a+x weka-3-*-6/jre/zulu17.32.13-ca-fx-jre17.0.2-linux_x64/lib/jspawnhelper

To see whether you have the correct permission settings, run:

ls -al weka-3-*-6/jre/zulu17.32.13-ca-fx-jre17.0.2-linux_x64/lib/jspawnhelper

The permissions should look like this:

-rwxr-xr-x

Cheers,
Eibe

PS: Regarding the Auto-WEKA package, we have submitted a pull request with some changes to the package source code that should ensure smooth operation when using Auto-WEKA with WEKA 3.8.6/3.9.6. A new release of the Auto-WEKA package will hopefully be available soon.

> On 8/02/2022, at 3:16 PM, Eibe Frank <[email protected]> wrote:
> 
> WEKA 3.8.6/3.9.6 has been released. The Windows, Mac, and Linux versions include a Java 17 runtime now. Several packages have also been updated accordingly (e.g., RPlugin and kfGroovy). Note that Auto-WEKA and the "distributed WEKA" packages still need some work to run with the Java 17 VM, but you can also run WEKA with a Java 8 VM if you have such a JVM installed separately on your system.
> 
> If you want to use WEKA from a system command-line interface (e.g., the Windows CMD or the macOS terminal), particularly with the included Java 17 VM, use the corresponding RunWeka.bat and weka.sh scripts included in the distribution. The former can be used to run the main method of any WEKA class by providing the "class" argument. Here is an example in the Windows Powershell:
> 
>  & 'C:\Program Files\Weka-3-9-6\RunWeka.bat' class weka.Run .MLRClassifier -t 'C:\Program Files\Weka-3-9-6\data\iris.2D.arff'
> 
> This uses the main() method in weka.Run to execute the MLRClassifier from the RPlugin package. When using weka.sh (on Linux and macOS), use "-main" instead of "class".
> 
> The Windows installer is code-signed now, for a slightly less scary installation process. The macOS application is code-signed as well now (please re-download the 3.8.6/3.9.6 release DMG if you encounter issues, we have updated the DMGs on SourceForge a couple of days ago to include code-signing).
> 
> The new version of WEKA comes with a more modern cross-platform look-and-feel by default, thanks to the fantastic Flatlaf project (https://www.formdev.com/flatlaf/). You can change this in the settings if you prefer another look-and-feel.
> 
> For the relatively small set of functionality and efficiency improvements in this release of WEKA, see
> 
>   https://www.cs.waikato.ac.nz/~ml/weka/CHANGELOG-3-9-6
> 
> The corresponding changes have also been ported to the 3.8.6 version.
> 
> For changes in the WEKA packages maintained by us, see https://www.cs.waikato.ac.nz/~ml/weka/CHANGELOG_PACKAGES-3-9-6. For example, there are new versions of RPlugin, netlibNativeLinux, kfGroovy, hotSpot, and associationRulesVisualizer. The RPlugin works on ARM Macs as well now. Taken together with the new netlibNativeOSXarm package, this means that support for this platform should be fairly complete (wekaDeeplearning4j excluded).
> 
> Cheers,
> Eibe 

_______________________________________________
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