Attributes ml.svm not working

irllabs / ml-lib

A machine learning library for Max and Pure Data

Other

274 stars 40 forks source link

Attributes ml.svm not working #130

Closed Jelkro closed 5 years ago

Jelkro commented 7 years ago

With Pd 0.46.7 on Ubuntu 16.04 LTS

Hi,

I'm working on a project that uses the ml.svm classifier. The classification works fine, however, some of the attributes that are displayed in the help file are not working correctly.

These return a "not supported" message:

load /tmp/model/
save /tmp/model/
normalise (was this perhaps replaced by "scaling"?)

Also, after setting parameter values of Cost and Nu, the values are shown to be unchanged when retrieving them with a getcost and getnu messages.

Since classification works fine with current settings, my main concern is that I would like to be able to save and load models.

Thanks!

jamiebullock commented 5 years ago

Hi @Jelkro I know this is a super old issue, and maybe you don't even use ml.lib any more, but in any case...

The correct messages for loading and saving the model are write and read. scaling 1 is indeed the correct message for normalisation.

As for cost and nu, oddly these only get set after the model is trained. This is a detail of libsvm. Maybe they get altered internally according to the training process, but in any case I didn't want to mess around with this.

HTH