Support #10357
closed
Support Vector Machines not working
Added by Gianpaolo Coro almost 8 years ago.
Updated almost 8 years ago.
Category:
High-Throughput-Computing
Infrastructure:
Production
Description
SVMs are not working on the DataMiner. If this depends on the conflict is in the libsmv inside RapidMiner, this can be deleted because we are not using the SVMs of RapidMiner.
I'm not the developer of the repidminer jar. I really don't know who is using it in the algorithms and if it can be changed.
Also this is a uberjar containing other jar, do you know how it is created?
No one is. RapidMiner is a third party framework. The current integrated version is 5, the last that was Open Source. The SVM library should be deleted either from the jar directly or from the RM clone here (a .jardesc executor is already available to generate the jar):
https://svn.d4science.research-infrastructures.eu/gcube/trunk/Common/RapidMiner/RapidMiner_Wasat
Alternatively, we can add it among the direct dependencies of DM, as it was previously. Anyway, we are talking about something that worked previously.
I add that RapidMiner is one of the most important frameworks integrated by the DataMiner (together with Weka), thus it should be legacy if possible.
It worked in the past only because the svn library integrated in the algortithm was the same version of the one in rapidminer.
So one way could be to make SVMs use the same lib version of RapidMiner?
- Status changed from New to Closed
- % Done changed from 0 to 100
We have switched to the SVMs of Rapid Miner which are equal to the one we were using but it is the "Svm" class instead of the "svm" class. The issue has been fixed.
Also available in: Atom
PDF
Updated by 
Updated by 
