R help - Sep 2009 - multiclass SVM (e1071 package): number of estimated models

Hi,

I run multiclass SVM for iris data, which contains 3 classes (manual page
52). Based on manual, the implementation uses one-against-one approach:
k*(k-1)/2 binary classifiers trained. However, I am getting only two models
instead of three (only two columns of support vectors and coefficients).
What do I miss?

Thanks a lot for help,
John

Bellow is the code.

package(e1071)
data(iris)
x <- subset(iris, select = -Species)
y <- Species
model <- svm(x, y)
model$SV
model$coefs
-- 
View this message in context:
http://www.nabble.com/multiclass-SVM-%28e1071-package%29%3A-number-of-estimated-models-tp25624020p25624020.html
Sent from the R help mailing list archive at Nabble.com.

Dear John,

there *are* indeed 3 classifiers trained, as you can see from

predict(model, iris, decision.values = TRUE)

However, the coefficients are stored in a compressed format -- see 
svminternals.txt in the /doc subdirectory.

Best
David

-----------------

I run multiclass SVM for iris data, which contains 3 classes (manual page
52). Based on manual, the implementation uses one-against-one approach:
k*(k-1)/2 binary classifiers trained. However, I am getting only two models
instead of three (only two columns of support vectors and coefficients).
What do I miss?