RM5 does not have some value series operators

awchisholm

Hello,

I just tried to use the RM5 value series capabilities and I noticed that the operators "MusicExampleSource" and "ValueSeriesGP" are missing.

Will they be coming soon?

best regards,

Andrew

land

Hi,
they will be released as Audio Processing Extension in the future, but we don't have a release date, yet. How serious do you need it?

Greetings,
 Sebastian

awchisholm

Hello Sebastian,

I'm learning and evaluating RM so it's not urgent or critical in any way for me. It would just make my life easier because I wouldn't have to use RM4.6 for some of the things I'm looking at and RM5 for others.

Could you perhaps release the VSGP operator in RM5 as part of the values series extension since it's a thing of power and usefulness beyond just audio.

regards

Andrew

land

Hi Andrew,
while the MusicExampleSource does not pose when migrating it to RapidMiner 5.0, especially the VSGP makes us serious headaches. We decided to remove it completely some month ago, because it does not meet our standards of stability we try to establish with RapidMiner 5.0. There would be needed a bunch of improvements to make this operator work that well, and since we did not thought it would be used anyway, we decided to drop it. It would be hard to migrate anyway and this work without a satisfying result wasn't worth the time. At least it wasn't worth the time for us.
But of course we are always willing to meet the needs of our users, especially if they are customers...

Greetings,
  Sebastian

awchisholm

Hello Sebastian,

I was using the VSGP output as a way of learning about value series in general. Some of the things it comes up with are interesting and instructive and yes there were one or two bugs but I could usually work around them. To practice, I did use it successfully to distinguish between spoken phrases. The accuracy I achieved was comparable to commercially available packages but the astonishing thing was that this was done without me having to understand the details of how.

I realised the approach could be applied to automatically extract features from any value series, not just audio data (of course, you knew that already  ) without me having to understand how. I was hoping to continue this with a much better idea I have (I'll make sure I buy the right license if I can make it work ).

It would be a shame not to have this operator in RM5 since I can't find anything like it in Knime, Orange or Weka to use (I don't know about commercial packages for financial reasons).

I subscribe to the "worse is better" approach; even things that appear bad to the developer can still be valuable to the user.

regards,

Andrew