R help - May 2011 - how to interpret coefficients from multiclass svm using libsvm (for multiclass R-SVM)

Hello all,

I'm working with the svm (libsvm) implementation from library(e1071).
Currently I'm trying to extend recursive feature elimination (R-SMV)  to
work with multiclass classification.

My problem is that if I run svm for a 3 class problem I get a 2-D vector
back from
model$coefs, can someone explain me what this values are? I understand them
in the 2-class problem
where this is  a 1-D vector...

the help only says :
coefs The corresponding coefficients times the training labels.


best regards
Immanuel

	[[alternative HTML version deleted]]